Implement decompression for vectored reads

[arpad-m]

Implement decompression of images for vectored reads.

This doesn't implement support for still treating blobs as uncompressed with the bits we reserved for compression, as we are removing that functionality in #8300 anyways.

Part of #5431

[github-actions]

3048 tests run: 2933 passed, 0 failed, 115 skipped (full report)

---

Flaky tests (3)

Postgres 15

• test_isolation[None]: debug
• test_tenant_creation_fails: debug

Postgres 14

• test_subscriber_restart: release

Code coverage* (full report)

• functions: 32.8% (6985 of 21326 functions)
• lines: 50.2% (54998 of 109620 lines)

* collected from Rust tests only

---

_{The comment gets automatically updated with the latest test results
c736123 at 2024-07-12T01:16:00.624Z :recycle:}

[jcsp]

Flaky test looks like something significant:

fixtures.pageserver.http.PageserverApiException: create image layers

[arpad-m]

Interesting, PS logs say (pageserver_3):

2024-07-08T19:47:17.192071Z ERROR request{method=PUT path=/v1/tenant/f897d0661d7a611c2f8a6e6d02db31d5-0204/timeline/0b988c50497f57d037bb39151ab4b352/checkpoint request_id=d626a796-7c25-4c2d-a8fd-404635bcf075}: Error processing HTTP request: InternalServerError(create image layers

Caused by:
    0: rename temporary file as correct path for 000000067F000035BA00000D7F0100000000-000000067F0000400000000A3A0300000000__00000000159A2270-00000001
    1: EEXIST: File exists

It uploaded that same file a little bit earlier as part of background compaction:

2024-07-08T19:46:54.286512Z  INFO compaction_loop{tenant_id=f897d0661d7a611c2f8a6e6d02db31d5 shard_id=0204}:compact_timeline{timeline_id=0b988c50497f57d037bb39151ab4b352}: scheduled layer file upload 000000067F000035BA00000D7F0100000000-000000067F0000400000000A3A0300000000__00000000159A2270-00000001 gen=00000001 shard=0204

I also see it upload the index_part.json after that message and before the erroring HTTP request:

2024-07-08T19:46:54.546389Z  INFO remote_upload{tenant_id=f897d0661d7a611c2f8a6e6d02db31d5 shard_id=0204 timeline_id=0b988c50497f57d037bb39151ab4b352 upload_task_id=90}: uploaded an index_part.json with future layers -- this is ok! if shutdown now, expect future layer cleanup disk_consistent_lsn=0/158E25D0

It's an interesting race. Maybe compression makes the situation worse due to taking more time, but in general it seems like an unrelated issue to me: there is earlier failures with the same error message (test_pg_regress[4] fails with a number of error messages).

[VladLazar]

Looks good. I would personally try a CI run of this PR rebased on #8288, but I leave it up to you.

[arpad-m]

I would personally try a CI run of this PR rebased on #8288

before enabling it in staging or prod, I certainly will do that. I'd like to handle any upcoming issues in followups.