Epic: stabilize Postgres prefetch

[ololobus]

Motivation

Konstantin and other people recently discovered several prefetch problems

DoD

There are no known prefetch problems. Existing issues fixed and covered with tests

Tasks

Tasks

Preview Give feedback

1. Prefetch doesn't work in Postgres 17
2. PG17 Konstantin's take Fix seqscan prefetch in pg17 #9865 (direct backport of <PG17 code)
3. PG17 Matthias' take Disable readstream's reliance on seqscan readahead #9860 (new PG17 flow)
4. Write a test with huge seqscan and comapre perf in <=PG16 and PG17 (new prefetch) -> seems good by Konstantin's tests
5. Index prefetch attemtps to prefetch invalid blocks #9866
   c/PostgreSQL t/bug +2
   
6. Fix issues with prefetch ring buffer resize #9847
   +0
7. Index-only-scan inconsistencies (??? hard to reproduce) -> upstream issues
8. Instability in test_prefetch #9961
   a/test +1
   
9. Set io_combine_limit=1 for PG 17

Other related tasks and Epics

• https://github.com/neondatabase/cloud/issues/12088

[ololobus]

This week:

• @knizhnik benchmark Matthias' PG17 prefetch against PG16 implementation Disable readstream's reliance on seqscan readahead #9860
• @knizhnik merge Fix issues with prefetch ring buffer resize #9847
• @knizhnik address comments in Check for invalid down link while prefetching B-Tree leave pages for index-only scan #9867, @MMeent review

[ololobus]

This week:

• @knizhnik @MMeent get PG 17 PR merged Disable readstream's reliance on seqscan readahead #9860
• @MMeent do another round of review for Check for invalid down link while prefetching B-Tree leave pages for index-only scan #9867

[ololobus]

This week:

• @MMeent merge Disable readstream's reliance on seqscan readahead #9860
• @MMeent do another round of review for Check for invalid down link while prefetching B-Tree leave pages for index-only scan #9867