fix(quality): skip provenance archives in brand-consistency scan

[neuron7xLab]

Problem

`tools/quality/brand_consistency.py` reports 66 violations across 4 file trees that contain legacy brand names (`TradePulse`, `HydroBrain`, `neurotrade`, `NeuroPro`) by design — they document the old-repo → GeoSync rebrand and lose audit value if scrubbed.

Fix

Added four directory prefixes to `SKIP_PREFIXES` (same pattern as existing `docs/archive`):

• `.claude/archive/` — rebrand provenance ledgers (ACTION_RESULT_ACCEPTOR_LEDGER, OLD_REPO_SALVAGE_LEDGER)
• `tests/archive/` — tests that exercise the archive ledger schemas
• `tools/archive/` — tooling that compares old-repo to new-repo by name
• `.claude/worktrees/` — ephemeral Claude Code agent worktrees, not repo source

Evidence

• Before: `✗ brand-consistency: 66 violation(s) across 3851 scanned files`
• After: `✓ brand-consistency: clean (5291 files scanned, 91 allowlisted)`
• Self-tests: 24/24 pass

What this is NOT

• Not adding individual allowlist entries — the prefix mechanism is the right tool when the entire subtree is provenance.
• Not weakening the gate — forbidden tokens still rejected outside these four prefixes.
• Not removing legacy names from archives — they remain audit evidence of the rebrand.

🤖 Generated with Claude Code