Rank Ghost Inspector failures by what to fix first

**Objective:** Produce a ranked triage list of the Ghost Inspector tests that failed in the last `{lookback-days}` days (default 7), optionally limited to the suite named `{suite-name}`, and recommend an order of work.

**Step 1: Narrow to a suite if one was given**

If `{suite-name}` is set, call `ghost-inspector-get-suite` with `name.contains` set to it. If more than one suite matches, list the matches and stop. Keep the suite `id`. If `{suite-name}` is empty, skip this step and triage everything.

**Step 2: Collect the failures**

Call `ghost-inspector-get-test-result` with `passing` false, `date-created.from` set to the start of the lookback window, `suite-id` from step 1 if you have one, and `max-records` set high enough to cover the window. If nothing comes back, report that no tests failed in the window and stop.

**Step 3: Collect the successes for the same window**

Call `ghost-inspector-get-test-result` again with the same filters but `passing` true. You need both sets to work out a failure rate rather than a raw count, because a test that runs hourly will always top a raw count.

**Step 4: Rank**

Group both sets by `test-name`. For each test compute runs, failures and failure rate. Rank by this order:

* tests failing on every run in the window, these are broken rather than flaky
* tests with a high failure rate and many runs
* tests that failed once at the end of the window, these may be a new regression
* tests that failed once in the middle and have passed since, lowest priority

Note which browsers and regions each failure occurred in. A test that only fails in one browser or one region is a different problem from one failing everywhere.

**Step 5: Report**

A table of: test name, suite, runs, failures, failure rate, browsers affected, most recent failure. Then a short ranked recommendation of what to investigate first and why, at most five items.

**Rules and guidelines:**

* Read only. Do NOT call `ghost-inspector-execute-test`, `ghost-inspector-execute-suite`, `ghost-inspector-update-test`, `ghost-inspector-update-suite`, `ghost-inspector-cancel-suite-run` or any create, duplicate or comment tool.
* Do not diagnose individual failures here. Point at them and say which to investigate.

**Validation:**

* If `{suite-name}` matches nothing, say so and offer to run across all suites instead.
* If the result count hits `max-records`, say the window was truncated and the ranking may be incomplete.