Shutting down no longer logs a warning for every view that had not yet drawn

A view that goes away before it ever draws anything is worth noticing: somebody was shown a spinner and then nothing. So the platform records a warning naming the view, which is what makes that situation findable instead of appearing as an unexplained delay.

But it recorded the same warning when the host itself was shutting down — and that is not the same event. When a host goes away, every view it serves is torn down at once, and any that had not yet drawn necessarily reports. A single ordinary shutdown therefore produced a burst of warnings, one per view, none of which indicated anything wrong.

Warnings are meant to be read. A recurring burst that never means anything trains people to skip them, which costs far more than the noise itself — the next real one is skipped too.

A view torn down because its host is shutting down is now recorded quietly, as routine lifecycle. A view abandoned for any other reason still warns, still names itself, and is as findable as before. This matches how the neighbouring case was already handled: a view that fails while its host is shutting down has been treated as routine for the same reason.