Add `corpus-report` command: readiness scorecard before first `write`

[telivity-otaip]

Why

New users ingest a messy LinkedIn dump and have no idea if the corpus is good enough for voice. A corpus-report (or status --verbose) that scores readiness drives successful first drafts and fewer "it sounds generic" issues.

Proposed command

personality-protect corpus-report
personality-protect corpus-report --json

Report sections (Contoso-safe aggregates only — never print piece text by default)

1. Counts by source (linkedin_post, linkedin_article, email, note, …)
2. Selection gate — how many pass select defaults; warn/block thresholds
3. Length bands — p50/p75/p90 words for posts vs articles
4. Date coverage — year histogram
5. Dedup hint — count of near-duplicates (reuse corpus_dedupe logic)
6. Article channel readiness — ≥5 articles in corpus + index?
7. Style card status — present / stale vs selection
8. Actionable next steps — e.g. "run index-voice", "ingest more articles"

Acceptance criteria

• Command exists and is listed in README CLI reference
• Default output is a Rich table/summary; --json emits machine-readable aggregates
• Never dumps raw personal prose unless --show-samples N on an explicit local flag (samples stay off by default; docs warn not to paste into issues)
• Unit tests with demo/synthetic fixtures
• status can link to or embed a short "run corpus-report for details" tip

Pointers

• select.py, style_profile.py, corpus_dedupe.py, cli.py status
• Demo profile via personality-protect demo

Difficulty

Easy–Medium — mostly composing existing metrics. Great for contributors learning the pipeline.