Make the @qe-style-checker comment trigger robust to surrounding prose

[mmcky]

Summary

The comment-trigger parser currently treats the entire comment body as @qe-style-checker <lecture> <categories>. Any extra prose in the comment breaks the parse, even when the command line itself is well-formed. The trigger should extract only the relevant command and ignore surrounding text.

Reproduction

A user posted this as an issue comment in QuantEcon/test-action-style-guide:

@qe-style-checker lectures/markov_chains_jax.md

But this has to be in a comment.

The run failed: test-action-style-guide run #27174490270

⚠️  Invalid categories: But
❌ Could not extract lecture name from comment
Process completed with exit code 1.

The command line @qe-style-checker lectures/markov_chains_jax.md was valid, but:

• the trailing word "But" was parsed as a rule category (invalid), and
• the multi-line prose caused lecture-name extraction to fail.

Expected behavior

A normal human comment that happens to contain the trigger should still work, e.g.:

@qe-style-checker lectures/markov_chains_jax.md — thanks!

or the trigger followed by explanatory text on later lines.

Suggested approach

• Parse only the line that contains @qe-style-checker, not the whole comment body.
• On that line, take the first token after the trigger as the lecture path; treat a following token as categories only if it matches the known category set (admonitions, code, figures, jax, links, math, references, writing) — otherwise ignore trailing tokens rather than erroring.
• Optionally: when categories are omitted, fall back to the default set instead of failing.
• Improve the error messages to echo the line that was parsed, so misuse is easier to diagnose.

Context

Found while running comment-triggered demos in the test repo. Current behavior forces users to post a bare command with no surrounding text, which is an easy footgun.