fix(cli): warn when review add --type is not a configured type

[martintrojer]

Fixes #681.

CommentType::from_id maps any non-empty id to Custom. That is right for the TUI, where ids come from validated comment_types config, but the CLI had nothing to check against — so --type isue was stored, exported as **[ISUE]**, and badged as an unknown type, all at exit 0.

Why warn instead of reject

Two reasons, and I would rather flag them than have them found in review.

comment_types is unset by default, and under that default every id but none is undefined. Rejecting unknown ids would make --type issue an error on a stock install — which is exactly what the agent skill tells agents to pass. Second, a scripted caller should not start failing because someone edited a config it does not control.

So: an absent or empty comment_types means no opinion and no warning. A configured list that omits the id gets a warning naming the valid ids. The comment is written either way and the exit code stays 0.

If you would rather have a hard error, or a --strict-type flag, say so and I will rework it — the placement is the same either way.

The warning goes to stderr, so stdout stays parseable JSON.

add already read config for the username fallback, so the check reuses that read rather than adding a second one; resolve_cli_author now takes the config instead of loading it.

Test plan

• cargo test: 1585 passed, 0 failed. cargo fmt --check and clippy --all-targets clean (one pre-existing bitbucket/bkt.rs warning, present on main)
• Five new unit tests: warns and names the valid ids for an unconfigured id, silent for a configured one, silent when comment_types is absent or the config is default, silent for none, and the comment is still written when the type is unconfigured
• Manually, with comment_types = [note, issue]:
  • --type issue → nothing on stderr
  • --type isue → Warning: comment type 'isue' is not configured; known types: note, issue on stderr, stdout still valid JSON (verified with json.load), exit 0
  • No config file: both --type issue and the none default stay silent

[agavra]

Thank you @martintrojer for the contribution! This looks good to me.