ref(opt): compile out Clay's unused debug-tools UI

[natemoo-re]

Clay's built-in debug inspector ships in our wasm but clayterm never uses it. Clay__RenderDebugView (~700 lines + ~150 UI string literals) is reachable from the exported Clay_EndLayout behind a runtime if (context->debugModeEnabled) branch, so --gc-sections can't strip it, and at -O2 clang inlines it into the layout monolith (the "55 KB function" that dominated twiggy). clayterm never calls Clay_SetDebugModeEnabled, so it's pure dead weight.

• Adds patches/clay-disable-debug-tools.patch: upstream-compatible #ifndef CLAY_DISABLE_DEBUG_TOOLS guards around the #pragma region DebugTools block and its single call site in Clay_EndLayout. The guards are harmless no-ops unless the flag is defined.
• Makefile applies the patch to the clay submodule at build time (idempotent — skips if already applied; reverted by make clean) and opts in via -DCLAY_DISABLE_DEBUG_TOOLS.
• raw wasm 155,934 → 111,234 bytes (−44,700, −28.7%)
• brotli (what ships) 41,809 → 30,417 bytes (−11,392)
• npm esm/wasm.js 53,143 → 38,903 bytes (−14,240, ≈ −13.9 KB unpacked) — wasm.js is the only changed package file
• Cold path only: debugModeEnabled is always false at runtime, so no executed layout code changes — none of the -Oz perf tradeoff. deno task test passes (8 / 146 steps); watching CodSpeed to confirm neutrality.
• Upstream opportunity: the region is already bracketed with #pragma region DebugTools / #pragma endregion, so this CLAY_DISABLE_DEBUG_TOOLS flag is a ~6-line PR to nicbarker/clay and would benefit every wasm Clay consumer. If accepted upstream, bump the submodule and drop the patch — only the -D flag remains.

[pkg-pr-new]

Open in StackBlitz

npm i https://pkg.pr.new/clayterm@89

commit: d08642c

[codspeed-hq]

Merging this PR will degrade performance by 13.64%

⚡ 1 improved benchmark
❌ 2 regressed benchmarks
✅ 20 untouched benchmarks

Warning

Please fix the performance issues or acknowledge them on CodSpeed.

Performance Changes

	Mode	Benchmark	BASE	HEAD	Efficiency
❌	Simulation	long input burst (200 bytes)	576.1 µs	1,264.5 µs	-54.44%
❌	Simulation	dashboard layout	439.9 µs	591.1 µs	-25.58%
⚡	Simulation	printable ASCII (single char)	174.5 µs	91.9 µs	+89.93%

Tip

Investigate this regression by commenting @codspeedbot fix this regression on this PR, or directly use the CodSpeed MCP with your agent.

---

_{Comparing ref/clay-debug-tools (d08642c) with main (48ff4bb)}