feat: Add lazy sinks

[coastalwhite]

This PR adds a lazy boolean flag to all sinks. If this is set to true, the sink returns a LazyFrame and .collect() needs to be called before it gets executed. The collect returns an empty DataFrame. This also now allows combination of sink_* and collect_all.

Example

import polars as pl
from pathlib import Path

p = Path("./")

lf = pl.LazyFrame({"a": [1, 2, 3]})
lf1 = lf.sink_parquet(p / "a.parquet", lazy=True)
lf2 = lf.sink_csv(p / "a.csv", lazy=True)

assert not Path(p / "a.parquet").exists()
assert not Path(p / "a.csv").exists()

pl.collect_all([lf1, lf2])

Fixes #6506.

[codecov]

Codecov Report

Attention: Patch coverage is 93.60465% with 11 lines in your changes missing coverage. Please review.

Project coverage is 81.02%. Comparing base (ea82623) to head (555af63).
Report is 1 commits behind head on main.

Files with missing lines	Patch %	Lines
crates/polars-lazy/src/frame/mod.rs	92.37%	9 Missing ⚠️
py-polars/polars/functions/lazy.py	66.66%	0 Missing and 2 partials ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main   #21733      +/-   ##
==========================================
- Coverage   81.03%   81.02%   -0.01%     
==========================================
  Files        1610     1610              
  Lines      233031   233003      -28     
  Branches     2685     2689       +4     
==========================================
- Hits       188837   188802      -35     
- Misses      43563    43570       +7     
  Partials      631      631              

☔ 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.

[thomasfrederikhoeck]

@coastalwhite why return the emtpy dataframe? Could that lead to misunderstandings with a query like:

pl.scan_csv("somefile").sink_csv("somefile2", lazy=True).collect().with_columns(pl.all() * 2).write_csv("somefile3")

Would it make more sense for .sink_ to not be a LazyFrame but is own type such as Sink which could maybe return the file paths instead (or similar) when called with .collect()?

[coastalwhite]

@coastalwhite why return the emtpy dataframe? Could that lead to misunderstandings with a query like:

pl.scan_csv("somefile").sink_csv("somefile2", lazy=True).collect().with_columns(pl.all() * 2).write_csv("somefile3")

Would it make more sense for .sink_ to not be a LazyFrame but is own type such as Sink which could maybe return the file paths instead (or similar)?

We could maybe do that. I would require some magic on the python side.

[thomasfrederikhoeck]

@coastalwhite the file-path was just a suggestion of something that could be meaningful to return. I think mostly I would be worried about the collect returning the empty DF.

[coastalwhite]

After thinking about it a bit more, for now, I don't think it is a good idea. If you explicitly say lazy, I am assuming you are going to do something with collect_all so it shouldn't matter that much.

[gab23r]

Fixed #6506

[ritchie46]

Fixed #6506

Almost the CSE part is incoming.