feat: impl UDTF for @udf

[KKould]

Extend @udf to enable it to return multiple rows of data.

@udf(
    input_types=["INT"],
    result_type=[("num", "INT"), ("double_num", "INT")],
    batch_mode=True,
)
def expand_numbers(nums: List[int]):
    import pyarrow as pa

    schema = pa.schema(
        [pa.field("num", pa.int32(), nullable=False), pa.field("double_num", pa.int32(), nullable=False)]
    )
    return pa.RecordBatch.from_arrays(
        [pa.array(nums, type=pa.int32()), pa.array([n * 2 for n in nums], type=pa.int32())],
        schema=schema,
    )