nix: apply patched wasm-opt (LSB-test → ctz) to RTS wasm files

[ggreif]

Summary

• Adds ggreif/binaryen@gabor/lsb-if-ctz-flake as a flake input, which exposes a patched wasm-opt implementing the optimization from WebAssembly/binaryen#8562
• In nix/rts.nix, runs wasm-opt --optimize-instructions on the three non-debug RTS variants during installPhase
• Debug variants are left unoptimized intentionally

The optimization

The patched wasm-opt folds LSB-test patterns common in the RTS into i32.ctz:

i32.const 1; i32.and; i32.eqz; br_if  →  i32.ctz; br_if
i32.const 1; i32.and; if T E           →  i32.ctz; if E T

These patterns appear 26–105 times per RTS variant, primarily in the is_skewed/is_scalar pointer-tagging checks in the GC hot path.

Benchmark results (EOP RTS, gross effect of --optimize-instructions)

Benchmark	Before	After	Δ
heap-32 (run 1)	1,153,792,735 instr	1,151,398,207 instr	−0.21%
heap-32 (run 2)	1,256,407,315 instr	1,253,408,059 instr	−0.24%
heap-64 (run 1)	1,324,057,357 instr	1,321,855,449 instr	−0.17%
heap-64 (run 2)	1,295,845,087 instr	1,293,744,743 instr	−0.16%
bignum	2,504,499 cycles	2,504,383 cycles	−116
candid-subtype-cost	1,115,011 cycles	1,114,823 cycles	−188

Numbers are gross effects (other rules in --optimize-instructions also contribute).

Notes

• Once WebAssembly/binaryen#8562 lands upstream, this can be switched to the official nixpkgs binaryen and the fork dependency dropped.

🤖 Generated with Claude Code

[github-actions]

Comparing from 18aedb0 to a1c0a71:
In terms of gas, 5 tests improved and the mean change is -0.0%.
In terms of size, 5 tests improved and the mean change is -0.3%.