pl.read_parquet regression in 0.15.4 + 0.15.5

[erinov1]

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

The example from #2654 breaks again on 0.15.5 and 0.15.5. This last worked with 0.15.2.

ArrowErrorException: OutOfSpec("validity mask length must match the number of values")

Reproducible example

import pandas as pd

df_temp = pd.DataFrame(
    {
        "a": [
            "V",
            "V",
            "V",
            "V",
            "V",
            "V",
            "V",
            "V",
            "V",
            "V",
            "V",
            "V",
            "V",
            "V",
            None,
            None,
            None,
            None,
            None,
            None,
        ]
    }
)
df_temp.to_parquet("a.parquet", index=False)
import polars as pl

pl.read_parquet("a.parquet")

Expected behavior

No regression in parquet reader.

Installed versions

---Version info---
Polars: 0.15.5
Index type: UInt32
Platform: macOS-10.15.7-x86_64-i386-64bit
Python: 3.8.12 (default, Oct 18 2021, 20:31:01) 
[Clang 10.0.1 (clang-1001.0.46.4)]
---Optional dependencies---
pyarrow: 10.0.0
pandas: 1.4.4
numpy: 1.21.5
fsspec: 2022.01.0
connectorx: 0.3.0
xlsx2csv: <not installed>
matplotlib: 3.6.2

[ritchie46]

fixed in #5797 will patch tonight.