fix: update native benchmark to use Fe4 field access
Fix a[0]/b[0] array indexing to a.l0/b.l0 in the unsignedMultiplyHigh
and uLt micro-benchmarks on Kotlin/Native linuxX64.
Native benchmark results (Kotlin/Native LLVM AOT, linuxX64):
verifySchnorr: 85,402 ns/op (11,709 ops/s)
verifySchnorrFast: 77,231 ns/op (12,948 ops/s)
signSchnorr: 74,702 ns/op (13,386 ops/s)
sign (cached pk): 41,273 ns/op (24,228 ops/s)
pubKeyCreate: 32,305 ns/op (30,954 ops/s)
ecdhXOnly: 100,152 ns/op (9,984 ops/s)
Field micro-benchmarks:
FieldP.mul: 44 ns/op FieldP.add: 5 ns/op
FieldP.sqr: 37 ns/op FieldP.sub: 6 ns/op
Batch verify: 3.1x (4 events) to 7.0x (32 events) faster
https://claude.ai/code/session_01Sxi6Gpxbstuj3Y8TBY7XrU
Claude
28b76f4320