feat(metrics): historical VCS metric trend (time series over N points)

[dekobon]

Follow-up to #328.

Add historical metric trend support: compute VCS (and optionally AST) metrics at N historical points in time, surfacing improving-vs-degrading code over the project's lifetime.

Why

A single snapshot of risk_score answers "what's risky now." A trend answers "is this getting better or worse" — which is the actionable question for technical-debt programs. The Kamei JIT survey notes that JIT models lose predictive power over time, which is a different framing of the same phenomenon: metrics decay; trends are more durable.

Scope

• New CLI flag: bca vcs trend --points <N> --span <DURATION> (e.g., 12 points over 24 months → monthly metric snapshots).
• For each point in time: check out the repo at that commit (or use gix to read trees without checking out), compute VCS metrics with --as-of set to that point.
• Emit a time-series JSON: { files: { path: [{ as_of, risk_score, ... }, ...] } }.
• Optional delta summary: which files improved / regressed the most.

Edge cases

• Files that didn't exist at an older time point: emit null.
• Renamed files: track via rename detection if --follow-renames.
• Very deep histories: cap point count; document.

Acceptance criteria

• Time-series output schema is stable and versioned.
• Performance: 12 monthly points on a 1k-commit fixture completes in under 60 s.