parquet_read fails when a column has too many rows with string values

[vincev]

The parquet_read and parquet_read_record fail to read a parquet file with a column of string values when the number of rows exceed a few thousand.

To reproduce the problem I generated the parquet file with the following script:

import sys
import pandas as pd
import numpy as np
import pyarrow as pa

print(f"Pandas version:  {pd.__version__}")
print(f"Numpy version:   {np.__version__}")
print(f"Pyarrow version: {pa.__version__}")

values = [f'{x:040}' for x in range(int(sys.argv[1]))]
df = pd.DataFrame({'values': values})
df.to_parquet('test.parquet', index=False, version='2.0')
print(f"Wrote {len(df)} rows")

With 20000 rows all works well:

> python gen.py 20000
Pandas version:  1.3.2
Numpy version:   1.21.2
Pyarrow version: 5.0.0
Wrote 20000 rows
> cargo run --release --example parquet_read test.parquet 0 0
    Finished release [optimized] target(s) in 0.12s
     Running `target/release/examples/parquet_read test.parquet 0 0`
Utf8[0000000000000000000000000000000000000000, 0000000000000000000000000000000000000001, 0000000000000000000000000000000000000002, 0000000000000000000000000000000000000003, 0000000000000000000000000000000000000004, ...(19990)..., 0000000000000000000000000000000000019996, 0000000000000000000000000000000000019997, 0000000000000000000000000000000000019998, 0000000000000000000000000000000000019999]

but with 30000 rows the example fails with NotYetImplemented:

> python gen.py 30000
Pandas version:  1.3.2
Numpy version:   1.21.2
Pyarrow version: 5.0.0
Wrote 30000 rows
> cargo run --release --example parquet_read test.parquet 0 0
    Finished release [optimized] target(s) in 0.11s
     Running `target/release/examples/parquet_read test.parquet 0 0`
Error: NotYetImplemented("Decoding \"Plain\"-encoded, dictionary-encoded optional V1 pages is not yet implemented for Binary")

I am running the latest master version:

> git rev-parse HEAD
cef5f08cf86334772d5ac72291f563e63c298e46