Improve double formatting performance on net8+ and fix equality incorrectness re special doubles

[mgravell]

by using double : IUtf8Formattable

open question: this is net8+; however, for anything netcoreapp2.1+ we could actually use the Span<char> overload to a scratch buffer, which would avoid a string alloc (and copy down if needed, like we do today from string). However, this would only really impact net6 which is EOL, so I do not propose to introduce extra code for that.

This PR also fixes a pre-existing oddity where, (as RedisValue):

• "na" != "NA"
• "nan" == "NAN"
• "nand" != "NAND"

(and similar for inf)

[NickCraver]

@mgravell I made you a present to use here: #2931

[mgravell]

delta, before (limiting to relevant) - note in particular allocs (scroll right):

| Method        | s      | value     | host     | port  | Mean       | Error     | StdDev    | Median     | Op/s          | Gen0   | Allocated |
|-------------- |------- |---------- |--------- |------ |-----------:|----------:|----------:|-----------:|--------------:|-------:|----------:|
| ParseDouble   | -1     | ?         | ?        | ?     |  56.424 ns | 1.1702 ns | 2.7122 ns |  56.113 ns |  17,722,894.2 |      - |         - |
| ParseDouble   | -inf   | ?         | ?        | ?     |   2.900 ns | 0.0993 ns | 0.1815 ns |   2.879 ns | 344,866,763.5 |      - |         - |
| FormatDouble  | ?      | -Infinity | ?        | ?     |   3.012 ns | 0.1001 ns | 0.1558 ns |   3.008 ns | 331,955,012.2 |      - |         - |
| FormatDouble  | ?      | -1        | ?        | ?     |  93.365 ns | 1.8845 ns | 2.8779 ns |  93.333 ns |  10,710,608.6 | 0.0050 |      32 B |
| ParseDouble   | 0      | ?         | ?        | ?     |   2.694 ns | 0.0872 ns | 0.1134 ns |   2.703 ns | 371,187,328.7 |      - |         - |
| FormatDouble  | ?      | 0         | ?        | ?     |  45.925 ns | 0.9217 ns | 0.9052 ns |  45.977 ns |  21,774,478.6 | 0.0038 |      24 B |
| ParseDouble   | 123442 | ?         | ?        | ?     |  81.211 ns | 1.6605 ns | 2.1000 ns |  81.056 ns |  12,313,592.5 |      - |         - |
| FormatDouble  | ?      | 123442    | ?        | ?     | 149.446 ns | 3.0354 ns | 4.8145 ns | 149.223 ns |   6,691,386.7 | 0.0062 |      40 B |
| ParseDouble   | 64     | ?         | ?        | ?     |  62.042 ns | 1.2612 ns | 2.5763 ns |  61.844 ns |  16,118,182.2 |      - |         - |
| FormatDouble  | ?      | NaN       | ?        | ?     |  13.124 ns | 0.2994 ns | 0.4920 ns |  13.071 ns |  76,197,072.3 |      - |         - |
| FormatDouble  | ?      | 64        | ?        | ?     | 124.475 ns | 2.5356 ns | 3.4708 ns | 124.267 ns |   8,033,736.3 | 0.0050 |      32 B |
| ParseDouble   | nan    | ?         | ?        | ?     |   3.468 ns | 0.1105 ns | 0.1512 ns |   3.459 ns | 288,367,071.9 |      - |         - |

after (note no allocs):

| Method        | s      | value     | host     | port  | Mean       | Error     | StdDev    | Op/s          | Gen0   | Allocated |
|-------------- |------- |---------- |--------- |------ |-----------:|----------:|----------:|--------------:|-------:|----------:|
| ParseDouble   | -1     | ?         | ?        | ?     |  54.380 ns | 1.1219 ns | 1.9051 ns |  18,389,178.2 |      - |         - |
| ParseDouble   | -inf   | ?         | ?        | ?     |   4.124 ns | 0.1213 ns | 0.2423 ns | 242,509,622.1 |      - |         - |
| FormatDouble  | ?      | -Infinity | ?        | ?     |   2.991 ns | 0.1011 ns | 0.2087 ns | 334,327,800.4 |      - |         - |
| FormatDouble  | ?      | -1        | ?        | ?     |  80.053 ns | 1.6459 ns | 4.2191 ns |  12,491,696.1 |      - |         - |
| ParseDouble   | 0      | ?         | ?        | ?     |   2.862 ns | 0.0978 ns | 0.1372 ns | 349,429,539.4 |      - |         - |
| FormatDouble  | ?      | 0         | ?        | ?     |  32.396 ns | 0.6835 ns | 0.9356 ns |  30,867,908.5 |      - |         - |
| ParseDouble   | 123442 | ?         | ?        | ?     |  77.273 ns | 1.5881 ns | 2.6094 ns |  12,941,150.5 |      - |         - |
| FormatDouble  | ?      | 123442    | ?        | ?     | 129.659 ns | 2.5727 ns | 5.5379 ns |   7,712,560.7 |      - |         - |
| ParseDouble   | 64     | ?         | ?        | ?     |  60.785 ns | 1.2505 ns | 3.0675 ns |  16,451,422.2 |      - |         - |
| FormatDouble  | ?      | NaN       | ?        | ?     |  20.836 ns | 0.4540 ns | 0.5047 ns |  47,994,833.7 |      - |         - |
| FormatDouble  | ?      | 64        | ?        | ?     | 109.433 ns | 2.1685 ns | 2.5814 ns |   9,138,011.9 |      - |         - |
| ParseDouble   | nan    | ?         | ?        | ?     |   3.551 ns | 0.1058 ns | 0.2114 ns | 281,591,155.7 |      - |         - |

only "regression" is parse -inf, from 2.9ns to 4.1ns - pretty sure that's just benchmark jitter /cc @NickCraver