Fix aggregations on a self-referencing if_else

[jaychia]

Describe the bug
Aggregations on self-referencing if_else is broken:

def test_repro():
    daft_df = DataFrame.from_pydict(GROUPBY_DATA).repartition(2)
    daft_df = daft_df.with_column(
        "weight", (col("size") == "S").if_else(1, col("weight"))
    )
    daft_df = daft_df.groupby(col("animal")).agg([(col("weight"), "mean")])
    daft_df.collect()

The above breaks with a missing ID during a projection step. It seems that the optimizer erroneously omits the "weight" column when doing column pruning.