flake: update nixpkgs input, nixVersions.nix_2_28 -> nixVersions.nix_2_30

[xokdvium]

Nix 2.30 has significanltly improved memory usage and eval performance,
which would be nice to have in nixd.

Unpinning boost182 is necessary because:

1. Now nix from nixpkgs propagates its boost build input.
2. Linking 2 different versions of boost into the same executable is rather risky
   wrt to ABI breaks.

[xokdvium]

Huh, UBSAN barfs on:

/nix/store/eeeeeeeeeeeeeeeeeeeeeeeeeeeeeeee-boost-1.87.0-dev/include/boost/asio/impl/thread_pool.ipp:133:7: runtime error: load of value 32, which is not a valid value for type 'bool'
SUMMARY: UndefinedBehaviorSanitizer: undefined-behavior /nix/store/eeeeeeeeeeeeeeeeeeeeeeeeeeeeeeee-boost-1.87.0-dev/include/boost/asio/impl/thread_pool.ipp:133:7 in

[xokdvium]

Bumping boost seems to help here, but some tests start failing for unclear reasons...

[inclyc]

Thanks for your contribution! It would be great if this patch could be merged, but the CI failures need to be fixed first.

[xokdvium]

The failures are really weird. When I run lit with a bunch of processes the regressions/nixd test suite fails, but that works when running the test suite with -j1 (passing it to lit).
This is cursed. I suppose something in nixpkgs broke this.

[xokdvium]

Ok, so the boost thing is broken because it really does have an uninitialized variable: boostorg/asio@30b5974

[xokdvium]

FML! This is a boost bug indeed.

@inclyc, the CI is now fixed. Here's what was causing those failures:

• Nix started to propagate the boost version, so the pinned boost182 was ignored by meson completely and it picked up 1.87 propagated by nix.
• boost::asio::thread_pool default constructor is broken in 1.87 (https://github.com/nix-community/nixd/pull/709/files#diff-a4cc3cdf892c650b5ba354c28bf468802c7353cc0d69ab4e59d6f03b9d26429aR136-R139) and doesn't initialize a boolean variable.

This caused either a) random failures b) UBSAN to spew errors (#709 (comment))

[inclyc]

LGTM, thank you!