feat: --project filter on issues list

[lightstrike]

Summary

Users can now filter issues by project when listing: lineark issues list --project "My Project". Accepts project name or UUID, and combines with existing filters like --team and --mine.

Also exposes comment UUIDs in issues read output, enabling the comments delete workflow.

Changes

• Add --project NAME-OR-ID flag to IssuesAction::List, using existing resolve_project_id helper
• Add id field to CommentSummary so comment UUIDs appear in issues read
• Update lineark usage reference

Test plan

• cargo clippy --workspace -- -D warnings — clean
• cargo fmt --check — clean
• cargo test --workspace — all passing
• lineark issues list --project "Project Name" --limit 5
• lineark issues list --project <uuid> --team ENG
• lineark issues list --help shows --project flag
• lineark issues read ENG-123 includes comment IDs

[flipbit03]

Functionally it LGTM, but I think this PR has a bit of "scope creep" that shouldn't be tackled here. Let's keep PRs focused on the mark. If some of the comments are really useless, we open up an issue to evaluate the value proposition of some of these comments, and then we tackle that work item separately.

[flipbit03]

The test additions look fine to me, but the comment removals are unrelated to the work being done, and I think some of these comments are contextually important enough in explaining crevices we explored during development - we shouldn't remove then without some planning.

[lightstrike]

Fair. I have some logic in my Claude harness to remove comments for "self-descriptive code" -- of course some subjectivity there. I'll revert.

[flipbit03]

@lightstrike PS: I wasn't expecting such an influx of PRs - Thanks a lot for your interest! To serve the project better and allow external PRs to be online-tested, I'm working to create a harness so that your PRs can also be tested against the online token after reviewing them. I think it's important that PRs test against an actual Linear workspace, to really confirm they work.

[lightstrike]

@flipbit03 Thanks for starting this project! It's a great foundation to build from. Great idea to have the harness in place and agreed it's important to make sure all of these work with real projects before merging in.

[flipbit03]

@lightstrike thanks for the tweaks! We can most surely address comment's signal-to-noise ratio in a further PR. issue opened in #97. I think you need to rebase this, and it's good to go

[flipbit03]

@lightstrike needs to be rebased on top of the current main. After that, just give me a ping and it's good to go! Thanks!

[lightstrike]

@flipbit03 Ready for review again — rebased onto main and reverted the unrelated comment deletions. All CI checks passing.

[flipbit03]

LGTM!

[flipbit03]

The new test fails online. I'll attempt to fix it - seems trivial

---- issues_list_with_project_filter ----
test panicked: workspace must have at least one project


failures:
    issues_list_with_project_filter

test result: FAILED. 49 passed; 1 failed; 0 ignored; 0 measured; 0 filtered out; finished in 49.52s

error: test failed, to rerun pass `-p lineark --test online`
Error: Process completed with exit code 101.

[flipbit03]

Hey @lightstrike — I pushed a fix for the failing test (two commits on the branch) - thanks for letting me push commits to your fork!

What happened: The issues_list_with_project_filter test assumed the workspace already had projects, but our CI test workspace is bare — so it panicked at "workspace must have at least one project".

What I changed:

• Rewrote the test to be self-contained: it now creates its own project and issue, uses RAII guards (ProjectGuard, IssueGuard) for automatic cleanup, and retry_with_backoff to handle Linear's eventual consistency.
• Added proper output.status.success() checks with stdout/stderr in assertion messages before parsing JSON — the original .unwrap() on JSON parsing was hiding the real error (e.g. an auth failure would show up as "EOF while parsing a value" instead of the actual 401 message).

Going forward: I've also added a section to CLAUDE.md documenting these online test conventions (self-contained tests, RAII guards, status checks before JSON parsing) so future contributors have clear guidance.

Confirmed passing locally against the live API. Should be good to merge now!