Add collect_column to dataframe

[timsaucer]

Which issue does this PR close?

Closes #1288

Rationale for this change

This PR adds a function to extract a single column of data from a dataframe and return it as a pyarrow array. It is based on user feedback that they must go through complicated lengths to get access to this data.

What changes are included in this PR?

• Adds collect_column function to DataFrame
• Adds unit test

Are there any user-facing changes?

This is a new API. No existing APIs have been modified.

[timsaucer]

@ntjohnson1 FYI

[ntjohnson1]

Ran through a couple usages I hit often and looks much nicer! Extracting a scalar as a python type is still a little verbose but dramatically better than before.

import datafusion as dfn
from datafusion import col
import pyarrow as pa

def main():
    ctx = dfn.SessionContext()

    df = ctx.from_pydict({
        "a": [3, 1, 2],
        "b": [4, 6, 5]
    })
    before_result = pa.table(df.select("a"))["a"].to_numpy()
    result = df.collect_column("a").to_numpy()
    print(f" {before_result=}\n\t{result=}")

    # Without as_py
    df_of_interest = df.filter(col("b") > 5)
    before_result = pa.table(df_of_interest.select("b"))["b"][0]
    result = df_of_interest.collect_column("b")[0]
    print(f" {before_result=}\n\t{result=}")

    # With as_py
    df_of_interest = df.filter(col("b") > 5)
    before_result = pa.table(df_of_interest.select("b"))["b"][0].as_py()
    result = df_of_interest.collect_column("b")[0].as_py()
    print(f" {before_result=}\n\t{result=}")

if __name__ == "__main__":
    main()

 before_result=array([3, 1, 2])
        result=array([3, 1, 2])
 before_result=<pyarrow.Int64Scalar: 6>
        result=<pyarrow.Int64Scalar: 6>
 before_result=6
        result=6