[FEAT] migrate schema inference → async, block at py boundary

[andrewgazelka]

Converts schema inference operations for CSV, JSON, and Parquet files to use async/await instead of synchronous runtime blocking. This architectural change ensures that blocking operations happen at the highest level possible (the Python API boundary) rather than deep within the inference logic. Key changes include:

• Making read_csv_schema, read_json_schema, and read_parquet_schema async
• Updating scan builder interfaces to use async finish() methods
• Removing unnecessary runtime.block_on calls from schema inference paths
• Moving runtime.block_on calls to Python API layer where blocking is unavoidable
• Converting schema-related tests to use tokio async runtime
• Adding common-runtime dependency where needed
• Fixes refactor: GlobScanOperator do not block runtime! #3423

This change improves the consistency of async IO handling and creates a cleaner architecture where blocking is consolidated at the Python interface rather than scattered throughout the codebase.

[andrewgazelka]

• [FEAT] migrate schema inference → async, block at py boundary #3432 👈 (View in Graphite)
• main

This stack of pull requests is managed by Graphite. Learn more about stacking.

[codspeed-hq]

CodSpeed Performance Report

Merging #3432 will degrade performances by 42.44%

_{Comparing andrew/async-glob (f23654a) with main (2db1233)}

Summary

❌ 1 regressions
✅ 16 untouched benchmarks

⚠️ Please fix the performance issues or acknowledge them on CodSpeed.

Benchmarks breakdown

	Benchmark	main	andrew/async-glob	Change
❌	test_show[100 Small Files]	23.2 ms	40.4 ms	-42.44%

[codecov]

Codecov Report

Attention: Patch coverage is 85.22727% with 26 lines in your changes missing coverage. Please review.

Project coverage is 77.36%. Comparing base (2db1233) to head (f23654a).
Report is 2 commits behind head on main.

Files with missing lines	Patch %	Lines
src/daft-csv/src/metadata.rs	58.62%	24 Missing ⚠️
src/daft-scan/src/glob.rs	91.66%	2 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #3432      +/-   ##
==========================================
+ Coverage   77.35%   77.36%   +0.01%     
==========================================
  Files         684      684              
  Lines       83637    83692      +55     
==========================================
+ Hits        64695    64751      +56     
+ Misses      18942    18941       -1     

Files with missing lines	Coverage Δ
src/daft-csv/src/python.rs	50.00% <100.00%> (+3.70%)	⬆️
src/daft-json/src/python.rs	49.15% <100.00%> (+3.69%)	⬆️
src/daft-json/src/schema.rs	75.94% <100.00%> (+0.70%)	⬆️
src/daft-parquet/src/python.rs	59.84% <100.00%> (+0.93%)	⬆️
src/daft-parquet/src/read.rs	75.29% <100.00%> (-0.05%)	⬇️
src/daft-scan/src/builder.rs	21.12% <100.00%> (+1.38%)	⬆️
src/daft-scan/src/lib.rs	60.34% <100.00%> (+0.06%)	⬆️
src/daft-scan/src/python.rs	77.04% <100.00%> (+0.31%)	⬆️
src/daft-sql/src/table_provider/read_csv.rs	91.25% <100.00%> (+0.22%)	⬆️
src/daft-sql/src/table_provider/read_parquet.rs	96.66% <100.00%> (+0.11%)	⬆️
... and 2 more

... and 2 files with indirect coverage changes