Move chunkwise aggregation into TSL part

[akuzm]

Its purpose is to enable vectorized aggregation, and it has to interact with DecompressChunk to build proper plans, both of which are TSL-only.

Disable-check: force-changelog-file

[codecov]

Codecov Report

Attention: Patch coverage is 89.33333% with 24 lines in your changes missing coverage. Please review.

Project coverage is 81.70%. Comparing base (59f50f2) to head (fe95431).
Report is 275 commits behind head on main.

Files	Patch %	Lines
tsl/src/chunkwise_agg.c	89.57%	2 Missing and 20 partials ⚠️
src/planner/planner.c	66.66%	0 Missing and 1 partial ⚠️
tsl/src/planner.c	90.90%	0 Missing and 1 partial ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #7159      +/-   ##
==========================================
+ Coverage   80.06%   81.70%   +1.63%     
==========================================
  Files         190      204      +14     
  Lines       37181    38015     +834     
  Branches     9450     9855     +405     
==========================================
+ Hits        29770    31060    +1290     
+ Misses       2997     2985      -12     
+ Partials     4414     3970     -444     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[svenklemm]

There are quite a few test changes where we no longer push down aggregates

[akuzm]

There are quite a few test changes where we no longer push down aggregates

Right, the Apache ones. I think it's OK, the purpose of implementing our own partitionwise aggregation was to support vectorized aggregation which is TSL-only anyway. And if people need chunkwise aggregation with Apache version, there is the standard PG optionenable_partitionwise_aggregate.

[svenklemm]

dont think enable_partitionwise_aggregate works anymore for hypertables