slice is not respected in the sense that apparently the whole file is loaded first. This results for me in OOM and a longer than expected time.

[zbenmo]

Checks

• I have checked that this issue has not already been reported.
• I have confirmed this bug exists on the latest version of Polars.

Reproducible example

%%time

df = (
    pl.scan_parquet('/kaggle/input/leash-BELKA/train.parquet')
    [157:169]
    .collect()
)

Log output

No response

Issue description

Trying to load specific rows from a big parquet file.

With parquet as above, I get OOM exception, after a long time passes.
With scan_csv of the matching file seems to work fine.

Expected behavior

As I indicate the slice, and it can be potentially used to optimize and read only a small portion of the file, I would be happy if this is indeed what happened.

Installed versions

--------Version info---------
Polars:               0.20.21
Index type:           UInt32
Platform:             Linux-5.15.133+-x86_64-with-glibc2.31
Python:               3.10.13 | packaged by conda-forge | (main, Dec 23 2023, 15:36:39) [GCC 12.3.0]

----Optional dependencies----
adbc_driver_manager:  <not installed>
cloudpickle:          2.2.1
connectorx:           <not installed>
deltalake:            <not installed>
fastexcel:            <not installed>
fsspec:               2024.2.0
gevent:               <not installed>
hvplot:               <not installed>
matplotlib:           3.7.5
nest_asyncio:         1.5.8
numpy:                1.26.4
openpyxl:             3.1.2
pandas:               2.2.2
pyarrow:              15.0.2
pydantic:             2.5.3
pyiceberg:            <not installed>
pyxlsb:               <not installed>
sqlalchemy:           2.0.25
xlsx2csv:             <not installed>
xlsxwriter:           <not installed>