fix: Layer delete on drop and eviction can outlive timeline shutdown

[koivunej]

This is a follow-up to #7051 where LayerInner::drop and LayerInner::evict_blocking were not noticed to require a gate before the file deletion. The lack of entering a gate opens up a similar possibility of deleting a layer file which a newer Timeline instance has already checked out to be resident in a similar case as #7051.

[koivunej]

Ideas on further refinement and/or abstracting the timeline dependency out: Whenever Layer needs a timeline, it should be accompanied by a guard; there is never any other situation.

[github-actions]

2496 tests run: 2376 passed, 0 failed, 120 skipped (full report)

---

Flaky tests (2)

Postgres 14

• test_wal_restore_initdb: debug
• test_wal_restore: debug

Code coverage* (full report)

• functions: 28.8% (7033 of 24442 functions)
• lines: 47.6% (43452 of 91368 lines)

* collected from Rust tests only

---

_{The comment gets automatically updated with the latest test results
e99be22 at 2024-03-11T09:33:55.649Z :recycle:}

[problame]

Is this really a follow-up to #7051 ?

Seems to me like it's a similar issue like what was fixed in #7051, but, for the eviction rather the download path.

Would like to see the PR description adjusted if you think that's the case.

[koivunej]

Is this really a follow-up to #7051 ?

Seems to me like it's a similar issue like what was fixed in #7051, but, for the eviction rather the download path.

Would like to see the PR description adjusted if you think that's the case.

Sorry, I had the auto-merge on. "Follow-up" as in "continuation of the story" or even the google definition "pursue or investigate something further" as in these cases I was thinking of only during weekend.