improve err msg on use of non-partitioned column

[marijncv]

Description

Updates the error message when trying to use a partition filter with a column that is not partitioned.

Code to trigger the error (adapted from #1219):

from datetime import date

import numpy as np
import numpy.random
import pandas as pd
from deltalake import DeltaTable, write_deltalake

nrows = 9
data = pd.DataFrame(
    {
        "year": np.repeat(2023, nrows),
        "month": np.repeat([1, 2, 3], nrows / 3),
        "values": numpy.random.normal(size=nrows),
    }
)
table_path = "tables/monthly"
write_deltalake(
    table_path,
    data,
    partition_by=["year", "month"],
    mode="overwrite",
)

nrows = 4
new_data = pd.DataFrame(
    {
        "year": np.repeat(2023, nrows),
        "month": np.repeat([4], nrows),
        "values": numpy.random.normal(size=nrows) + 10,
    }
)

write_deltalake(
    table_path,
    new_data,
    mode="overwrite",
    partition_filters=[("values", "=", "1")],
)
DeltaTable(table_path).to_pandas()

Related Issue(s)

closes #1218

Documentation

[houqp]

nitpick, the original implementation returns all columns that aren't partitioned, but this implementation does an early return, which results in a different behavior.

[marijncv]

Thanks @houqp, good point! I tried to restore to the original behavior in the latest commit

[wjones127]

Sorry it took a bit to get to reading this. Looks good :)