SkipScan over compressed chunks

[natalya-aksman]

PR implements https://github.com/timescale/eng-database/issues/693 via "SkipScan <- DecompressChunk <- Compressed Index" plan.

SkipScan consumes DecompressChunk data to get the next distinct value, then uses it to update SkipScan qual on the compressed index under DecompressChunk node.

Current unit test setup: skip_scan_load creates an extra table skip_scan_htc with the same setup as skip_scan_ht but with compressed data. Then the same correctness and plan tests run on skip_scan_ht are also run on skip_scan_htc. Correctness checks pass, plan tests only apply to scenarios with index on "dev".

TODO:

• Create separate skip_scan_compressed unit test for tables with data compressed in various ways.
• Create separate unit tests for distinct aggregates over compressed data.
• Address SkipScan scenarios with DISTINCT on (dev), time ... ORDER BY dev, time on compressed data with segmentby=dev, orderby=time with ASC/DESC and NULLS LAST/FIRST for time.
• Enable SkipScan for indexes with NULL direction not matching segmentby NULL direction if distinct column is guaranteed to be NOT NULL, for example if it's a distinct aggregate input, or there is a NOT NULL constraint on this column or NOT NULL predicate is pushed down into the index. Address it as a separate issue: [Enhancement]: Enable SkipScan for indexes with NULL direction not matching guaranteed not-NULL distinct column direction #7996
• Improve cost model of SkipScan over compressed data

[svenklemm]

Have you done any benchmarking for this?

[natalya-aksman]

Have you done any benchmarking for this?

Not yet but existing SkipScan benchmark in tsbench should work as it already tests same queries on uncompressed vs. compressed data, so compressed data runs should speed up dramatically.

[codecov]

Codecov Report

Attention: Patch coverage is 94.24460% with 8 lines in your changes missing coverage. Please review.

Project coverage is 82.24%. Comparing base (59f50f2) to head (452baa4).
Report is 1009 commits behind head on main.

Files with missing lines	Patch %	Lines
tsl/src/nodes/skip_scan/planner.c	94.28%	1 Missing and 5 partials ⚠️
tsl/src/nodes/skip_scan/exec.c	93.10%	1 Missing and 1 partial ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #7983      +/-   ##
==========================================
+ Coverage   80.06%   82.24%   +2.17%     
==========================================
  Files         190      253      +63     
  Lines       37181    46896    +9715     
  Branches     9450    11807    +2357     
==========================================
+ Hits        29770    38570    +8800     
- Misses       2997     3652     +655     
- Partials     4414     4674     +260     

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

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[natalya-aksman]

Have you done any benchmarking for this?

Done it now, it shows 2000x-2500x faster runs for SkipScan queries on ht_metrics_compressed, see https://grafana.ops.savannah-dev.timescale.com/d/fasYic_4z/compare-benchmark-runs?orgId=1&var-branch=All&var-run1=4324&var-run2=4327&var-threshold=0.02&var-use_historical_thresholds=true&var-threshold_expression=2%20%2A%20percentile_cont%280.90%29&var-exact_suite_version=true