Replace hard-coded "id" with column_id in roll_time_series

[david-waterworth]

This fixes the inconsistency of _roll_out_time_series aways naming the resulting tuple index as "id" even if the user passes and alternative id via the column_id parameter.

[david-waterworth]

I think now that I've made this change, the return from roll_time_series also needs to change - the map_reduce may no longer contain an "id" column yet the return always sorts by "id"

I'll also add tests before this is ready to merge as well.

shifted_chunks = distributor.map_reduce(
    _roll_out_time_series,
    data=range_of_shifts,
    chunk_size=chunksize,
    function_kwargs=kwargs,
)

distributor.close()

df_shift = pd.concat(shifted_chunks, ignore_index=True)

return df_shift.sort_values(by=["id", column_sort or "sort"])

[nils-braun]

I think now that I've made this change, the return from roll_time_series also needs to change - the map_reduce may no longer contain an "id" column yet the return always sorts by "id"

Correct - you would want to also change this to column_id.

Thanks for your effort!