[PERF] scan task in memory estimate

[samster25]

• Closes: Allow daft.read_parquet take in custom statistics from the user #1898

1. When column stats are provided, use only the columns in the materialized schema to estimate in memory size

• when column stats are missing, fall back on schema estimate for that field

2. When num_rows is provided, use the materialized schema to estimate in memory size
3. When neither are provided, estimate the in memory size using an inflation factor (same as our writes) and approximate the number of rows. Then use the materialized schema to estimate in memory size
4. thread through the new in memory estimator to the ScanWithTask physical op

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 85.54%. Comparing base (1a94752) to head (0a1cdc8).

❗ Current head 0a1cdc8 differs from pull request most recent head 6bafa32. Consider uploading reports for the commit 6bafa32 to get more accurate results

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #1901      +/-   ##
==========================================
+ Coverage   83.93%   85.54%   +1.61%     
==========================================
  Files          55       55              
  Lines        6112     6228     +116     
==========================================
+ Hits         5130     5328     +198     
+ Misses        982      900      -82     

Files	Coverage Δ
daft/execution/rust_physical_plan_shim.py	93.93% <100.00%> (-1.52%)	⬇️

... and 15 files with indirect coverage changes

[jaychia]

Seems good, should we write some unit tests for these?