Improved flexibility of reading parquet

[jorgecarleitao]

This exposes an API to read parquet in different ways:

• read (IO-bounded) columns in parallel (!= concurrently)
• read and deserialize columns one by one (reduce memory pressure for column-based execution engines)

This allows the both of the two worlds: allow reading both multiple fields and distributed/concurrent work over individual fields.

[codecov]

Codecov Report

Merging #820 (8539846) into main (4fbbd90) will decrease coverage by 0.00%.
The diff coverage is 82.35%.

@@            Coverage Diff             @@
##             main     #820      +/-   ##
==========================================
- Coverage   71.27%   71.27%   -0.01%     
==========================================
  Files         327      327              
  Lines       17544    17543       -1     
==========================================
- Hits        12504    12503       -1     
  Misses       5040     5040              

Impacted Files	Coverage Δ
src/io/parquet/read/file.rs	69.38% <ø> (ø)
src/io/parquet/read/row_group.rs	74.24% <82.35%> (-0.39%)	⬇️

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 4fbbd90...8539846. Read the comment docs.