Streaming join after scan_parquet failing

[mishpat]

Polars version checks

• I have checked that this issue has not already been reported.

• I have confirmed this bug exists on the latest version of Polars.

Issue description

Streaming join only ouputs a small fraction of the output when scanning from a parquet file.
If this is done on an in-memory lazy frame it works fine, e.g. lf0.collect().lazy().join(lf1.collect().lazy(), ...).collect(streaming=True).

Reproducible example

import numpy as np
import polars as pl

np.random.seed(111)

n = 500_000
k = 100
d0 = {f"x{i}": np.random.random(n) for i in range(k)}
d0.update({"id": np.arange(n)})
df0 = pl.DataFrame(d0)
df1 = df0.clone().select(pl.all().shuffle(111))
print("lf0 shape:", df0.shape)
df0.write_parquet("df0.parquet")
df1.write_parquet("df1.parquet")

lf0 = pl.scan_parquet("df0.parquet")
lf1 = pl.scan_parquet("df1.parquet").select(pl.all().suffix("_r"))  # haven't merged the fix to this yet
df2 = lf0.join(lf1, left_on="id", right_on="id_r").collect(streaming=True)
print("post-streaming join shape:", df2.shape)

lf0 shape: (500000, 101)
post-streaming join shape: (816, 201)

Expected behavior

post-streaming join shape: (500000, 201)

Installed versions

--------Version info---------
Polars:      0.18.4
Index type:  UInt32
Platform:    Windows-10-10.0.19044-SP0
Python:      3.10.12 | packaged by conda-forge | (main, Jun 23 2023, 22:34:57) [MSC v.1936 64 bit (AMD64)]

----Optional dependencies----
numpy:       1.25.0
pandas:      <not installed>
pyarrow:     12.0.1
connectorx:  <not installed>
deltalake:   <not installed>
fsspec:      <not installed>
matplotlib:  <not installed>
xlsx2csv:    <not installed>
xlsxwriter:  <not installed>