tests: run the brigade skill's ticket add and fire lines against the tool

[creedants]

A new test runs the ticket add and fire command lines the brigade skill shows, so an example cannot drift from what brigade.py accepts without the suite failing.

• tests/test_skill_commands.py reads t3/added/brigade/SKILL.md, fills each placeholder in the alternative it runs, and runs every line against a scratch store. A line the tool refuses fails the test and names the line number in the skill.
• The test fails, not skips, when it finds fewer than seven command lines.
• Three regression tests run on a fixed sample held in the module, so a valid edit to a live example does not fail them.
• docs/guide.md now names edits.tsv among the tables open writes.

Known limits. The scanner reads physical lines and does not follow a continued shell command. A bare command mention with no flags is skipped. Only the five-kind choice runs every alternative; other choices run their first. Tilde fences are not recognized.

Reviewed at 52c106e5.