report: count obfuscator.io corpus in readability metrics

[vasie1337]

What

The synthetic javascript-obfuscator (obfuscator.io) fixtures in samples/generated/ — 140 files = 20 seeds × 7 obfuscation profiles — gated correctness via manifest.json, but were excluded from the readability metrics. Both the live report binary and the committed SCOREBOARD.md read samples/ non-recursively, so the obfuscator.io samples never showed up in the dashboard.

This adds a per-profile rollup of the generated corpus (aggregated over all seeds, one row per obfuscation technique) to both surfaces, so the obfuscator.io samples count toward readability the same way they count toward correctness.

Metrics shape

• kept% — byte-weighted Σout/Σin (lower = more dead weight removed)
• opaque% — mean of per-file opaque-name ratios
• rounds — worst case across the profile's files
• converged — yes only if every file reached a fixpoint

profile	files	kept%	brackets	opaque%
minimal	20	62%	12	8%
strarr_base64	20	68%	12	5%
strarr_rc4	20	72%	12	4%
controlflow	20	32%	12	8%
deadcode	20	35%	12	8%
numbers_keys	20	53%	12	5%
strong	20	57%	12	5%
all	140	56%	84	6%

Changes

• src/bin/report.rs — append an "obfuscator.io corpus, aggregated by profile" section to the live dashboard.
• tests/golden.rs — matching per-profile rollup in render_scoreboard(), so the committed table and live dashboard stay in agreement.
• tests/snapshots/SCOREBOARD.md — re-blessed with the new section.

Verification

• Live report runs and prints the new section.
• scoreboard_is_current passes, and re-runs deterministically without blessing.
• cargo clippy clean.

https://claude.ai/code/session_01EjhNTCU89wa5zaeRHMnfEc

---

Generated by Claude Code