fix(ci): rebuild from clean in size-report so build-config changes are measured

[natemoo-re]

The size report builds head then base without cleaning in between. For PRs that change only build configuration (Makefile, patches/, compiler flags) and not src/*.c, make sees clayterm.wasm as newer than the unchanged sources and skips the base rebuild — so the base step re-measures the head artifact, the delta comes out 0, and no size comment is posted even though the wasm size actually changed.

• Run make clean && make for both the head and base builds so each is a from-scratch compile.
• Surfaced by ref(opt): compile out Clay's unused debug-tools UI #89 (compile out Clay's debug-tools UI, ≈ −14 KB unpacked): the reduction is real in CI but the report can't see it because that PR only touches Makefile / patches/. Once this lands and ref(opt): compile out Clay's unused debug-tools UI #89 re-runs, its size comment will populate.

[pkg-pr-new]

Open in StackBlitz

npm i https://pkg.pr.new/clayterm@90

commit: 96efff6

[codspeed-hq]

Merging this PR will degrade performance by 40.53%

❌ 2 (👁 2) regressed benchmarks
✅ 21 untouched benchmarks

Performance Changes

	Mode	Benchmark	BASE	HEAD	Efficiency
👁	Simulation	long input burst (200 bytes)	576.1 µs	1,265.1 µs	-54.46%
👁	Simulation	diff render (second frame)	690.4 µs	888.9 µs	-22.33%

---

_{Comparing fix/size-report-rebuild (96efff6) with main (48ff4bb)}