Add SILK frame orchestration and public EncodeSILK API

[thomas-vilte]

Description

this closes out the SILK encoder split from #147 — the last piece. It's the orchestrator that wires everything already merged in internal/silk/ into a full frame encode (VAD → pitch analysis → noise shaping → prediction coefs → NSQ → range-coded bitstream, in decode order) plus a public EncodeSILK method on the root Encoder

also lands the useInterpolatedNLSFs fix flagged as a known gap in #165:
findLPCNLSF 's interpolation search now only runs when complexity >= 4 , matching silk_setup_complexity in libopus. The root Encoder already had a complexity field wired to CELT, so this just threads it through to the SILK encoder too

EncodeSILK is a separate entry point from Encode / EncodeFloat32 , not an automatic CELT/SILK switch — bitrate based auto-selection still always picks CELT. This is for callers who specifically want a SILK-only voice packet

also applies a DC-block high-pass to EncodeSILK 's input ( applySILKDCBlock, reusing the same fixed 3Hz filter CELT's path already had from #137) — libopus applies this to the shared PCM ahead of both CELT and SILK, and EncodeSILK was skipping it entirely. The pitch-adaptive VoIP cutoff variant ( hp_cutoff in libopus) is not implemented; this only ports the fixed-cutoff dc_reject path

not implemented (disclosed, follow-up work): the rate-control bisection loop, LBRR, and DTX. This port is single-pass, same philosophy as the CELT encoder series (correct simple v1 first)

round-trip tested end to end: encode with both the internal silk.Encoder and the public EncodeSILK , decode with the real decoder, for all three SILK bandwidths (NB/MB/WB)

Reference issue

part of the #147 split. Closes out the split entirely — last piece

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 92.07%. Comparing base (aa7cd85) to head (486cf3f).

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #166      +/-   ##
==========================================
+ Coverage   90.85%   92.07%   +1.22%     
==========================================
  Files          54       55       +1     
  Lines        9576     9792     +216     
==========================================
+ Hits         8700     9016     +316     
+ Misses        638      547      -91     
+ Partials      238      229       -9     

Flag	Coverage Δ
go	92.07% <100.00%> (+1.22%)	⬆️

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

☔ View full report in Codecov by Harness.
📢 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.