v0.9.0+: ARM NEON SIMD support

[kolkov]

Description

Add ARM NEON SIMD optimizations for Apple Silicon and ARM servers.

Status: ⏸️ DEFERRED

Waiting for Go native SIMD support (golang/go#73787) expected in Go 1.26+.

Current State

• x86-64 AVX2/SSSE3 SIMD implemented in assembly
• ARM falls back to pure Go implementation

When Go SIMD lands

• Migrate x86-64 SIMD from assembly to Go native SIMD
• Add ARM NEON using Go native SIMD
• Benchmark on Apple M1/M2/M3
• Benchmark on ARM servers (AWS Graviton)

References

• simd/archsimd: architecture-specific SIMD intrinsics under a GOEXPERIMENT golang/go#73787 - Go native SIMD proposal
• Current SIMD: simd/memchr_amd64.s, prefilter/teddy_ssse3_amd64.s

Milestone

v0.9.0+ (after Go 1.26)