feat(ingest,context): --type and --tag on ingest, --tag filter on context

[GQAdonis]

Replaces #33, which closed when its stacked base branch was deleted at merge; rebased onto main (contains #31).

Summary

Plan PR A3 of the team-aware learning work (skill-pack design §2, §8 U3).

• pk ingest --type <T> sets the OKF entry type (Lesson, Gotcha, Decision, …). The librarian used to force Reference on every compiled entry. That is now only the default.
• pk ingest --tag <t> is repeatable and adds tags such as role:api-dev or vis:team.
• Both are applied through the new Librarian::compile_with(raw, CompileOverrides) after the model's compile result. Caller tags come first and duplicates are dropped. compile() delegates to it, so existing callers are unchanged.
• pk context --tag <t> is repeatable and all-of: an entry must carry every requested tag. The filter runs before scoring.
• The pk-mcp knowledge_search and knowledge_ingest tools are out of scope, because the hooks use the CLI.

Test plan

• cargo test -p pk-cli --test tags_and_type runs the real CLI end to end against a local OpenAI-compatible endpoint that returns fixed compile results:
  • three ingests produce Lesson + role:api-dev, Gotcha + role:ui-dev + vis:team, and an untyped entry that stays Reference;
  • context --tag role:api-dev returns exactly the api entry;
  • two tags match only the entry that carries both;
  • a non-matching combination returns nothing;
  • with no tags, all three come back.
• cargo test -p pk-cli --test context_scoring (A1) still passes.

🤖 Generated with Claude Code

[chatgpt-codex-connector]

You have reached your Codex usage limits for code reviews. You can see your limits in the Codex usage dashboard.
To continue using code reviews, add credits to your account and enable them for code reviews in your settings.