PG16 - Propagate statistics without a user-specified name

[naisila]

Relevant PG commit:
postgres/postgres@624aa2a

Check only last commit for now, until we merge PG16 support for regression tests.

[codecov]

Codecov Report

Merging #7127 (64c7c90) into main (682dca1) will decrease coverage by 0.02%.
The diff coverage is 95.55%.

❗ Current head 64c7c90 differs from pull request most recent head e619127. Consider uploading reports for the commit e619127 to get more accurate results

@@            Coverage Diff             @@
##             main    #7127      +/-   ##
==========================================
- Coverage   93.40%   93.38%   -0.02%     
==========================================
  Files         273      273              
  Lines       58956    58995      +39     
==========================================
+ Hits        55068    55093      +25     
- Misses       3888     3902      +14     

[onderkalaci]

Just noting our internal discussion here: Giving auto-names is a tricky task, especially with sharding + partitioning involved. The reason is that there might be name collusions, and those are non-trivial to spot.

Although the proposed code changes might work as-is, we decided to deprioritize this patch for now, and focus on robustness over new PG16 features. We might come back to this once we resolve all the other PG16 integration issues.

@naisila maybe we should mark this as a draft for now?