Superseed `filter_by_sf()` with `filter_by_spatial_extent()` and create `crop_gtfs()`

[dhersz]

filter_by_trip_geometry() would work exactly like filter_by_sf().

filter_by_stop_location() would, on the other hand, filter only based on stop locations. Given a bbox/polygon, the function would keep/drop the stops located inside the polygon and the entries related to them.

Both of them would receive a sf object, which would make the name filter_by_sf() a bit ambiguous (arguably it doesn't communicate the function behavior very well even considering the current behavior alone, as shown by the several issues that have been opened addressing this function). So I think the best way forward is to soft deprecate filter_by_sf() in the release, warning that filter_by_trip_geometry() should be used instead, and to hard deprecate it in version 2.0.0.

[dhersz]

Ok, maybe the names are not the best.

Usually in filter_by_* functions the * is exactly what you have to supply to the function. E.g. filter_by_trip_id() receives trip_id and subsets the content based on that. In the case of filter_by_trip_geometry() and filter_by_stop_location() the user wouldn't pass the trip geometries and the stop locations. Rather, they would pass a spatial object and the filter behavior would change based on the function.

So, perhaps, to make the function behavior clearer we could rename filter_by_sf() to filter_by_spatial_extent() and what would be filter_by_stop_location() to crop_gtfs(). Feels better.