WASM: Intrinsify numberOfLeadingZeros, numberOfTrailingZeros, and bitCount

[lax1dude]

I implemented intrinsic functions for numberOfLeadingZeros, numberOfTrailingZeros, and bitCount using the clz, ctz, and popcnt instructions in the WASM and WASM GC backends.

The classlib unit tests pass and my app still works, I make use of the leading and trailing zero functions in my direct buffer malloc/free implementation so I actually do have a reason to optimize them.

[konsoletyper]

I doubt it will work with optimizations enabled. Optimizer may inline method bodies, written in Java, which makes impossible to apply these intrinsics. So additionally Wasm GC BE should provide class transformers which turns intrinsic target methods into native and removes their body.

[lax1dude]

@konsoletyper I see what you mean, I'll fix it tomorrow

[konsoletyper]

Yes, and although I won't ask you to write unit tests, please, at least disassemble your app and make sure that relevant WebAssembly instructions are used instead of falling back to "software" implementations.

[lax1dude]

@konsoletyper Okay I added the class transformers. I also added unit tests for the 64-bit numberOfLeadingZeros and numberOfTrailingZeros because I noticed they didn't exist. I also copied the code from the unit tests and compiled it myself and disassembled the binary and all the intrinsics are compiling correctly (on the AGGRESSIVE optimization level).