develop gtfs cleaner to handle repeated pair (trip_id, departure_time) warnings

[ethan-moss]

Description of the Feature to be Added

Calling is_valid() regularly leads to repeated pair (trip_id, departure_time) in stop_times.txt but these are not cleaned by clean_feed(). Further investigation shows that:

• typically this is because neighbouring rows in stop_times.txt, within a given trip, have duplicate trip_id, arrival_time, departure_time, and stop_id. It's clear this circumstance is an erroneous duplication.
• less often, there are neighbouring stops very close together, but they have different stop_ids - it's possible for these to be timetabled with the same time because of their proximity. Here is an example where a bus turns around and serves two stops on either side of the road (near Dingestow):
  

(OPTIONAL) Suggested Implementations

In stop_times.txt, when records have duplicate trip_id, arrival_time, departure_time, and stop_id these can be safely dropped (represent the same trip, stop and time). Develop a clean which drops these from the stop_times feed, to prevent these warnings from remaining.

Additional context

[r-leyshon]

@ethan-moss
I need help in understanding the issue:

1	typically this is because neighbouring rows in stop_times.txt, within a given trip, have duplicate trip_id,
arrival_time, departure_time, and stop_id. It's clear this circumstance is an erroneous duplication.

2	less often, there are neighbouring stops very close together, but they have different stop_ids - it's possible for
these to be timetabled with the same time because of their proximity. Here is an example where a bus turns around
and serves two stops on either side of the road (near Dingestow):

Point 1 seems straight forward - whenever consecutive trips have the same trip_id, arrival_time, departure_time, and stop_id, drop the duplicate.

Point 2 - I am uncertain if it asks for any intervention. 2 different stops that are very close together and have the same arrival / departure time (not specified above). As these have different stop_ids they would not qualify for cleaning in point 1. Therefore I am uncertain as to whether this point calls for any treatment. Could you let me know if I have misunderstood the ask?

[r-leyshon]

Migrated to datasciencecampus/assess_gtfs#5