chore: prepare release v0.9.23

[kolkov]

Summary

• ARM64 NEON GEMM + element-wise ops (feat: ARM64 NEON SIMD kernels (GEMM, element-wise ops) #106, feat(cpu): ARM64 NEON GEMM micro-kernel and element-wise ops (#106) #152)
• WebGPU subgroup shaders: MatMul, BatchMatMul, Softmax (perf(webgpu): use subgroupShuffleXor for parallel reduction in compute shaders #141, perf(webgpu): subgroup cooperative shaders for MatMul, BatchMatMul, Softmax (#141) #151)
• ADR-019: opaque backend data (ADR-019: Remove GPU knowledge from core tensor (opaque backend data) #126, refactor: ADR-019 — remove GPU knowledge from core tensor (#126) #149)
• CPU hot paths + SIMD threshold (perf: CPU hot paths still doing per-element index math (gather, reduce, scatter-add, batch-broadcast) #81, perf(simd): add minimum slice-length threshold before dispatching to SIMD #90, perf(cpu): eliminate hot path index math + add SIMD threshold (#81, #90) #150)
• CHANGELOG/ROADMAP updated

Test plan

• go build ./... + ARM64 + WASM
• go test -short ./... — 22/22 pass
• golangci-lint run — 0 issues
• gofmt — clean

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.

📢 Thoughts on this report? Let us know!