The fault hunt files one issue for each cause degraded more than ten minutes in its day, quoting the cause's sentence #1297

For P0 Trust, under fail fast and loud (#1270) and the sentinel (#1131). This is the last of three pieces split out of #1272. It waits on #1296 for the history, and on #1131, which builds the hunt this rule plugs into.

Done looks like

• [ ] The hunt reads spans(db, dayStart, dayEnd) and files one finding for each cause whose spans total more than ten minutes in the day.
• [ ] Each finding follows #1131's rule. It is one issue carrying the cause, the sentence, the first open time, the total minutes and the href, labelled P0 Trust. If that cause already has an open finding, the hunt comments on it instead of filing a second issue.
• [ ] Spans under ten minutes in total are not filed. They stay in the history.
• [ ] Tests on in-memory SQLite, given planted spans: a 25-minute queue span files one issue; the same span on the next night comments on that issue; a 4-minute span files nothing.

Files: the sentinel's module from #1131, with its test.

Risks: filing twice is the failure to guard against. The test that the next night comments rather than files is the one that matters.

Brief: whoever builds #1131 (scout). Assign nobody until #1296 and #1131 land.