pageserver: coalesce index uploads when possible

[erikgrinaker]

Problem

With upload queue reordering in #10218, we can easily get into a situation where multiple index uploads are queued back to back, which can't be parallelized. This will happen e.g. when multiple layer flushes enqueue layer/index/layer/index/... and the layers skip the queue and are uploaded in parallel.

These index uploads will incur serial S3 roundtrip latencies, and may block later operations.

Touches #10096.

Summary of changes

When multiple back-to-back index uploads are ready to upload, only upload the most recent index and drop the rest.

[github-actions]

7304 tests run: 6933 passed, 0 failed, 371 skipped (full report)

---

Flaky tests (1)

Postgres 17

• test_lr_with_slow_safekeeper: release-arm64

Code coverage* (full report)

• functions: 32.9% (8146 of 24779 functions)
• lines: 48.1% (68268 of 141935 lines)

* collected from Rust tests only

---

_{The comment gets automatically updated with the latest test results
17120eb at 2025-01-14T20:29:17.955Z :recycle:}

[VladLazar]

Coalesced indices are always in the same generation, so this feels pretty safe.
Coalescing won't kick in until the flush loop stops inserting barriers though, right?

[erikgrinaker]

Coalescing won't kick in until the flush loop stops inserting barriers though, right?

That's right.

[erikgrinaker]

Coalesced indices are always in the same generation, so this feels pretty safe.

Seems worth making that explicit, I'll add a condition.

[erikgrinaker]

Coalesced indices are always in the same generation, so this feels pretty safe.

Seems worth making that explicit, I'll add a condition.

Can't check this, since IndexPart doesn't have the generation, but it must be true since a timeline instance has a static generation. Added a comment.