Pin ppx_deriving fork with forwarding optimizations

[sim642]

Issue #265.

The pinned fork optimizes the problems mentioned in #265 (comment) by omitting eta-expansion and unit arguments in simple cases, where they are unnecessary. I also opened a PR to ppx_deriving (ocaml-ppx/ppx_deriving#252) to hopefully get these optimizations upstream, allowing us to get rid of the pin.

On the added benchmarks (https://github.com/goblint/analyzer/blob/a21f33511183074c693c15269990210b283ae045/bench/deriving/benchEq.ml), this improves the performance of derived equal 3 times, bringing it more or less up to par with a completely manual implementation:
(deriving(_module)? is the new one, deriving(_module)?_expand is the old one)

*** Run benchmarks for path "module"

Throughputs for "manual", "deriving", "deriving_expand", "deriving_expand_simpl1", "deriving_expand_simpl2", "deriving_expand_simpl3" each running for at least 1 CPU second:
                manual:  1.24 WALL ( 1.19 usr +  0.01 sys =  1.20 CPU) @ 1721503707.76/s (n=2061015226)
              deriving:  1.02 WALL ( 1.00 usr +  0.00 sys =  1.00 CPU) @ 1312165613.22/s (n=1318050676)
       deriving_expand:  1.03 WALL ( 1.03 usr +  0.00 sys =  1.03 CPU) @ 319989570.35/s (n=329818050)
deriving_expand_simpl1:  1.07 WALL ( 1.06 usr +  0.00 sys =  1.06 CPU) @ 1316607444.91/s (n=1397158805)
deriving_expand_simpl2:  2.62 WALL ( 2.62 usr +  0.00 sys =  2.62 CPU) @ 1313578735.67/s (n=3441060051)
deriving_expand_simpl3:  1.11 WALL ( 1.11 usr +  0.00 sys =  1.11 CPU) @ 1254004556.80/s (n=1394683804)
                               Rate deriving_expand deriving_expand_simpl3 deriving deriving_expand_simpl2 deriving_expand_simpl1 manual
       deriving_expand  319989570/s              --                   -74%     -76%                   -76%                   -76%   -81%
deriving_expand_simpl3 1254004557/s            292%                     --      -4%                    -5%                    -5%   -27%
              deriving 1312165613/s            310%                     5%       --                    -0%                    -0%   -24%
deriving_expand_simpl2 1313578736/s            311%                     5%       0%                     --                    -0%   -24%
deriving_expand_simpl1 1316607445/s            311%                     5%       0%                     0%                     --   -24%
                manual 1721503708/s            438%                    37%      31%                    31%                    31%     --
**********************************************************************
*** Run benchmarks for path "pair.fst"

Throughputs for "manual_primitive", "deriving_primitive", "manual_module", "deriving_module", "deriving_module_expand", "deriving_module_expand_simpl1", "deriving_module_expand_simpl2" each running for at least 1 CPU second:
             manual_primitive:  1.33 WALL ( 1.32 usr +  0.00 sys =  1.32 CPU) @ 1043208945.28/s (n=1381680174)
           deriving_primitive:  1.14 WALL ( 1.14 usr +  0.00 sys =  1.14 CPU) @ 1391755623.57/s (n=1591025802)
                manual_module:  1.06 WALL ( 1.06 usr +  0.00 sys =  1.06 CPU) @ 1243419689.31/s (n=1318778383)
              deriving_module:  1.13 WALL ( 1.13 usr +  0.00 sys =  1.13 CPU) @ 1560610803.69/s (n=1755868185)
       deriving_module_expand:  1.09 WALL ( 1.09 usr +  0.00 sys =  1.09 CPU) @ 240938154.63/s (n=262628612)
deriving_module_expand_simpl1:  1.09 WALL ( 1.08 usr +  0.00 sys =  1.08 CPU) @ 284771156.32/s (n=308329135)
deriving_module_expand_simpl2:  1.08 WALL ( 1.12 usr +  0.00 sys =  1.12 CPU) @ 1397085184.51/s (n=1558836913)
                                      Rate deriving_module_expand deriving_module_expand_simpl1 manual_primitive manual_module deriving_primitive deriving_module_expand_simpl2 deriving_module
       deriving_module_expand  240938155/s                     --                          -15%             -77%          -81%               -83%                          -83%            -85%
deriving_module_expand_simpl1  284771156/s                    18%                            --             -73%          -77%               -80%                          -80%            -82%
             manual_primitive 1043208945/s                   333%                          266%               --          -16%               -25%                          -25%            -33%
                manual_module 1243419689/s                   416%                          337%              19%            --               -11%                          -11%            -20%
           deriving_primitive 1391755624/s                   478%                          389%              33%           12%                 --                           -0%            -11%
deriving_module_expand_simpl2 1397085185/s                   480%                          391%              34%           12%                 0%                            --            -10%
              deriving_module 1560610804/s                   548%                          448%              50%           26%                12%                           12%              --
**********************************************************************
*** Run benchmarks for path "pair.snd"

Throughputs for "manual_primitive", "deriving_primitive", "manual_module", "deriving_module", "deriving_module_expand", "deriving_module_expand_simpl1", "deriving_module_expand_simpl2" each running for at least 1 CPU second:
             manual_primitive:  1.13 WALL ( 1.13 usr +  0.00 sys =  1.13 CPU) @ 438248508.20/s (n=495223882)
           deriving_primitive:  1.08 WALL ( 1.07 usr +  0.00 sys =  1.07 CPU) @ 442271278.58/s (n=474449610)
                manual_module:  1.05 WALL ( 1.04 usr +  0.00 sys =  1.04 CPU) @ 445978757.51/s (n=466018367)
              deriving_module:  1.07 WALL ( 1.07 usr +  0.00 sys =  1.07 CPU) @ 439682085.65/s (n=468568759)
       deriving_module_expand:  1.06 WALL ( 1.05 usr +  0.00 sys =  1.05 CPU) @ 109149441.96/s (n=114901945)
deriving_module_expand_simpl1:  1.05 WALL ( 1.05 usr +  0.00 sys =  1.05 CPU) @ 163268054.55/s (n=171139534)
deriving_module_expand_simpl2:  1.07 WALL ( 1.07 usr +  0.00 sys =  1.07 CPU) @ 444435898.99/s (n=476877053)
                                     Rate deriving_module_expand deriving_module_expand_simpl1 manual_primitive deriving_module deriving_primitive deriving_module_expand_simpl2 manual_module
       deriving_module_expand 109149442/s                     --                          -33%             -75%            -75%               -75%                          -75%          -76%
deriving_module_expand_simpl1 163268055/s                    50%                            --             -63%            -63%               -63%                          -63%          -63%
             manual_primitive 438248508/s                   302%                          168%               --             -0%                -1%                           -1%           -2%
              deriving_module 439682086/s                   303%                          169%               0%              --                -1%                           -1%           -1%
           deriving_primitive 442271279/s                   305%                          171%               1%              1%                 --                           -0%           -1%
deriving_module_expand_simpl2 444435899/s                   307%                          172%               1%              1%                 0%                            --           -0%
                manual_module 445978758/s                   309%                          173%               2%              1%                 1%                            0%            --

[sim642]

The ppx_deriving PR was merged upstream, so that's good, but there's no release yet which contains it. So currently ppx_deriving is still pinned but to upstream, until they release it.