[CHORE] Expose read_sql partition bound strategy and default to min-max

[colin-ho]

Currently, read_sql calculates partition bounds using the PERCENTILE_DISC function. However, this function does not scale well to large tables, as it is an expensive window + sort function. A better alternative is to take samples, then estimate partition bounds, as described in this issue: #3245.

In the meantime, we should default to using the min-max calculations instead, which was previously the fallback option.

[codspeed-hq]

CodSpeed Performance Report

Merging #3246 will improve performances by 50.05%

_{Comparing colin/fix-readsql-partition-bounds (4fea9d7) with main (6e28b3f)}

Summary

⚡ 1 improvements
✅ 16 untouched benchmarks

Benchmarks breakdown

	Benchmark	main	colin/fix-readsql-partition-bounds	Change
⚡	test_show[100 Small Files]	50.1 ms	33.4 ms	+50.05%

[codecov]

Codecov Report

Attention: Patch coverage is 16.66667% with 35 lines in your changes missing coverage. Please review.

Project coverage is 78.52%. Comparing base (2b71ffb) to head (4fea9d7).
Report is 19 commits behind head on main.

Files with missing lines	Patch %	Lines
daft/sql/sql_scan.py	14.63%	35 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #3246      +/-   ##
==========================================
- Coverage   79.13%   78.52%   -0.62%     
==========================================
  Files         640      641       +1     
  Lines       77983    79095    +1112     
==========================================
+ Hits        61715    62107     +392     
- Misses      16268    16988     +720     

Files with missing lines	Coverage Δ
daft/io/_sql.py	51.85% <100.00%> (ø)
daft/sql/sql_scan.py	25.54% <14.63%> (-0.02%)	⬇️

... and 34 files with indirect coverage changes

[colin-ho]

Oops i forgot to mark this as ready @desmondcheongzx :P

[desmondcheongzx]

Looks good!