Find the GitHub commits that may have broken a Ghost Inspector test

**Objective:** Find the code changes most likely to have broken the Ghost Inspector test named `{test-name}`, searching back `{lookback-days}` days (default 14).

**Step 1: Find the test (Ghost Inspector)**

Call `ghost-inspector-get-test` with `name.contains` set to `{test-name}`. If nothing matches, report that and stop. If more than one test matches, list them with their suite name and id and stop so the user can choose.

**Step 2: Find the break point (Ghost Inspector)**

Call `ghost-inspector-get-test-result` with `test-id` set to the test's id and `date-created.from` set to `{lookback-days}` days ago. Order the results by `date-execution-started` yourself. Find the earliest failing result that is not followed by a pass (the first failure of the current failing streak) and the last passing result before it.

* If the test is currently passing, say so and stop.
* If there is no passing result in the window, report that the test has failed for the whole window, suggest a longer `{lookback-days}`, and continue using the start of the window as the lower bound.

**Step 3: Carry the window across (GitHub)**

The window is from the last passing result's `date-execution-started` to the first failing result's `date-execution-started`. Call `github-get-commit` with `commit-date.from` and `commit-date.to` set to those two times. Then call `github-get-workflow-run` with `created-date.from` and `created-date.to` set to the same window to see which builds and deploys ran.

Ghost Inspector and GitHub may report times in different time zones. Compare them in UTC, and if either side has no zone, widen the window by one hour on each side and say that you did.

**Step 4: Report**

* The test: name, suite, id, and the last-pass and first-fail times.
* Commits in the window: sha (short), author, date and first line of the message, newest first.
* Workflow runs in the window: name, branch, conclusion, created date and url.
* A short judgement of which commits are the strongest suspects, based on whether the message mentions the page, feature or URL the test covers (its `start-url` and name). Label this as inference.

**Rules & Guidelines:**

* `ghost-inspector-get-test` and `ghost-inspector-get-test-result` are on the ghost-inspector server; `github-get-commit` and `github-get-workflow-run` are on the github server.
* This prompt is read-only. Do NOT call `ghost-inspector-execute-test`, `ghost-inspector-add-test-result-comment`, `github-add-issue` or `github-add-issue-comment`.

**Validation:**

* If no commits fall in the window, say so plainly: the cause is more likely data, environment or the test itself than code, and point the user at the failing result's steps and console log.
* Never claim a commit caused the failure. Report correlation only.