perf(compute): add ARM64 NEON arithmetic kernels

[fallintoplace]

Summary

• Added ARM64 NEON kernels for primitive add, subtract, multiply, absolute value, and negate.
• Covered int32, uint32, int64, uint64, float32, and float64 values.
• Kept the generic path for unsupported widths and 64-bit integer multiply.
• Added edge cases for vector tails, wrapping arithmetic, minimum integers, signed zero, and the no-assembly fallback.

Benchmark

Apple M1 Pro, 3 MiB input, no nulls, median of 3 runs from BenchmarkScalarArithmetic:

Case	ARM64 NEON	-tags noasm	Speedup
int64 add, array-array	179,747 ns/op	759,116 ns/op	4.22x
int64 add, array-scalar	160,435 ns/op	697,497 ns/op	4.35x
float64 add, array-array	182,029 ns/op	761,779 ns/op	4.19x
float64 add, array-scalar	174,644 ns/op	714,281 ns/op	4.09x

Tests

• go test ./arrow/compute/... -count=1
• go test -tags noasm ./arrow/compute/... -count=1
• go test -race ./arrow/compute/internal/kernels ./arrow/compute -count=1
• go vet ./arrow/compute/internal/kernels
• Linux ARM64 and AMD64 cross-builds for the touched packages