Commit web JS codegen outputs in src/, verified by check-c-api

[bwhitman]

Why

amy_api.generated.js and patches.generated.js were generated into the gitignored build/ by make web, so consumers that commit copies (tulipcc's amyboardweb static/) could silently drift from their pinned amy — that's exactly what happened in shorepine/tulipcc#1208 after #943 removed voices.

What

• Move gen_amy_js_api.py and gen_patches_js.py outputs to committed src/amy_api.generated.js / src/patches.generated.js (alongside amy_c_api.generated.js from Table-driven cross-platform C API binding codegen #942), and give both scripts a --check mode.
• New scripts/gen_pcm_presets_js.py → committed src/pcm_presets.generated.js: the web editor's wave-preset table (window.AMY_WAVE_PRESETS) from pcm_tiny.h. This file previously existed only as a one-off committed in tulipcc with no generator anywhere; output is byte-identical to it apart from the header comment.
• make c-api runs all the generators; make check-c-api (already run by CI on every PR) now fails if any of the three JS outputs is stale.
• web/deploy-web use the committed files; the build/ rules are gone.

Testing

• make check-c-api passes; make test passes (117 tests).
• tulipcc side (PR to follow): amyboardweb stops committing static/*.generated.js copies and copies these from src/ at build time, so a bare pin bump can never drift them again.

🤖 Generated with Claude Code

[github-actions]

🎛️ AMY HW CI (AMYboard bench)

Flashed this PR's AMY (LoadTestChord: 6-voice Juno patch=1, one held note every 2 s) onto the physical AMYboard and measured the smoothed render load as the chord grows — back-to-back with the same sketch built at the PR's merge base, so Δ is this PR's own cost.

✅ PASS — the bench ran the test to completion.

notes held	main @ 85bfc77	this PR	Δ
1	1386	1386	+0
2	1546	1552	+6
3	2136	2139	+3
4	2327	2328	+1
5	2974	2972	-2
6	3169	3169	+0

Full chord settled render μs: 3170 (was 3171, Δ -0.0%) (peak 3181, 39 samples)

⬇️ Artifacts: serial log · load trace · report

_{Self-hosted bench (amyboardci). FAIL means only that the test could not run — the load values are informational, with no threshold and no audio compare. See tools/arduino_loadsweep/.}

[bwhitman]

⛓️ tulipcc integration PR opened

This merge was pinned into tulipcc for full-system CI: shorepine/tulipcc#1212

Test it there and merge that PR to move tulipcc onto this AMY.