- Full FFT overlap-add IR convolution in IRLoader (process(), set_mix(), toggle)
- Lazy FFT computation — IR FFT padded to correct block+ir size on first process()
- Wet/dry mix control, enabled/disabled toggle with tail clearing
- Fixed pipeline._apply_ir_cab() to delegate to IRLoader.process() instead of
poking internals (old code had array-size mismatch bug: IR FFT at ir_len vs
block FFT at conv_size)
- 46 tests: loading, convolution correctness, overlap-add state, mix, toggle,
directory listing, performance budget (all <5ms even at 8192 taps), edge cases
- scripts/download_irs.sh: free IR pack downloader (God's Cab, Seacow)
- 15 FX blocks implemented with per-block state isolation
- All blocks <500us per 256-sample block (reverb closest at 465us on x86)
- 57 unit tests all passing (per-effect, chain, bypass, state isolation)
- Benchmark script at scripts/benchmark_fx.py
- Vectorised delay reads via read_block_varying()
- scipy.lfilter for EQ (3-band RBJ) and reverb damping
- Fixed _DelayLine.write_block wraparound crash for large blocks
- Comb/Allpass buffers sized to BLOCK_SIZE * 2 minimum
shawn