Stopping a reindex in its first seconds is ignored and leaves the run stuck as Running #35182

Problem

If someone stops the Search Indexing application in the first few seconds of a run, while the reindex is still being set up, the stop is not honored:

• The reindex still runs. A few seconds after the stop, the server picks the job up again and indexes everything the user asked to stop.
• The run's status goes back to Running and never reaches a final state. It shows as Running until the next reindex starts.

In Collate, Admin Ops treats a Running app as busy. So after this happens, every new reindex request is refused with a 409, and the Admin Ops job that tracked the stopped run stays Running for up to an hour. This is what has kept Collate CI red on main since 9 October (AdminOpsApiIT.testCancelReindexJobMarksFailed and testReindexJobEndsWhenItsRunIsStopped).

Expected behaviour

A stop is honored however early it arrives: nothing is indexed after it, and the run ends as Stopped.

Acceptance criteria

• Stopping a run during its setup ends it as Stopped, and it stays Stopped.
• No entity is indexed for that run after the stop.
• A new reindex can be started right after.

How to reproduce

Trigger SearchIndexingApplication, then call POST /v1/apps/stop/SearchIndexingApplication about one second later, before Reindexing job started appears in the server log. GET /v1/apps/name/SearchIndexingApplication/runs/latest shows stopped briefly, then running, and never ends. The server log shows Restored appRunRecord to running a few seconds after the stop.

Seen in Collate CI runs 37938923627 and 38031946718.