🧹 Replace Unsafe.As<bool, byte> with (isWhite ? 1 : 0)

[eduherminio]

Benchmarks were not conclusive back in the day about this alleged perf. improvement

In @tannergooding words, Unsafe.As<bool, byte>(ref isWhite) can be notably less efficient than isWhite ? 1 : 0 plus
it technically opens the door to undefined behavior (since bool isn't strictly 0 or 1).
Regarding the benchmarks:

a lot of this differences are within 1-20 cycles (roughly 5-10ns)
and they are basically impossible to measure with traditional benchmarking tools
but, the JIT is set up to recognize and optimize the idiomatic patterns
so even if the raw difference is small, the potential difference in broader code can be massive
as it can make the difference on whether things like dead code elimination or constant folding occur

Having unsafe code that doesn't bring any benefits doesn't make sense anyway, and this regression test proved that removing the unsafe solution doesn't lose significant elo, so.. merging

Test  | perf/remove-unsafe-cast
Elo   | 0.38 +- 1.87 (95%)
SPRT  | 8.0+0.08s Threads=1 Hash=32MB
LLR   | 2.94 (-2.25, 2.89) [-3.00, 1.00]
Games | 56200: +15557 -15496 =25147
Penta | [1320, 6574, 12272, 6593, 1341]
https://openbench.lynx-chess.com/test/934/

[vitorhnn]

I'm sorry if this has been previously benchmarked, but why not isWhite ? 0 : 6?

[eduherminio]

You're obviously 100% right @vitorhnn, the BenchmarkDotNet benchmark actually tests that.
The only plausible explanation is that I opened this PR at 2 AM 😅