Summary

Closes #81 — the 3 residual correctness/quality follow-ups from PR #75 that #77/#78/#79 did not address.

Changes

• R1 EditType::from_db_str no longer distorts unknown DB values into AddSkill: added EditType::Unknown, made "add_skill" explicit (the old wildcard caught it), and _ => Unknown. Unknown is excluded from all() so it never pollutes edit-type coverage.
• R2 edit_type_roundtrips test timestamp is now index-based (enumerate), not as_str().len() — structurally unique, survives a future UNIQUE constraint or new variant.
• R3 token_estimate / estimate_tokens documented as intentional scaffold (not silently unused) — kept populated for future planner cost-weighting.

Acceptance Criteria

• AC1 ✅ from_db_str_unknown_is_not_distorted_to_add_skill + unknown_excluded_from_all (new tests).
• AC2 ✅ round-trip test timestamp is structurally unique (index-based).
• AC3 ✅ token_estimate/estimate_tokens carry scaffold doc rationale; clippy clean.
• AC4 ✅ cargo test --lib 651 passed; clippy clean; fmt clean.

Test Plan

• cargo test --lib: 651 passed
• clippy --all-targets -D warnings: clean
• cargo fmt --check: clean
• CI green