[1/3] Reduce SILK decode allocations

[zshang-oai]

Summary

Reduce avoidable allocation churn in the SILK decoder hot path. This is part of a series PRs to get an 2x performance boost with low hanging fruits.

What changed

• reuse decoder-owned scratch buffers for SILK header flags, pulse/excitation data, LPC working buffers, residual buffers, NLSF/LPC coefficient work, pitch lags, and LTP filter coefficients
• skip LBRR discard decoding when the packet does not signal any LBRR frames
• reuse previousFrameLPCValues storage instead of allocating a fresh slice on each handoff
• replace tiny pulse partition scratch allocations with stack arrays

Why

The SILK decode path was allocating transient slices on nearly every packet. In local stress benchmarks those allocations were a large avoidable part of the Pion/Opus gap.

This PR keeps the decode algorithm unchanged and only removes repeated scratch allocation work.

Validation

go test ./...

[codecov]

Codecov Report

❌ Patch coverage is 96.55172% with 3 lines in your changes missing coverage. Please review.
✅ Project coverage is 82.56%. Comparing base (de54ba8) to head (906eb7b).

Files with missing lines	Patch %	Lines
internal/silk/decoder.go	96.55%	1 Missing and 2 partials ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #114      +/-   ##
==========================================
- Coverage   82.96%   82.56%   -0.41%     
==========================================
  Files          22       22              
  Lines        4697     4742      +45     
==========================================
+ Hits         3897     3915      +18     
- Misses        609      635      +26     
- Partials      191      192       +1     

Flag	Coverage Δ
go	82.56% <96.55%> (-0.41%)	⬇️

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[Copilot]

Pull request overview

This PR reduces allocation churn in the SILK decoder hot path by reusing decoder-owned scratch buffers and avoiding unnecessary LBRR discard work when no redundancy is signaled.

Changes:

• Adds reusable scratch storage to Decoder for header flags, pulse/excitation buffers, LPC/NLSF work buffers, pitch/LTP data, and stereo-related decode state.
• Refactors allocation-heavy helpers to resize and clear decoder-owned buffers instead of allocating fresh slices.
• Short-circuits LBRR discard decoding when no mid/side LBRR flags are present.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.