Maintenance/gtfs network refactor w unit tests

[sablanchard]

Maintenance update with no substantial API changes comprised of code simplification, docstring and print updates, and unit test coverage expansion focusing on gtfs.network.py which include:

• updates to gtfs.network.py:

  • refactor save_processed_gtfs_data() and load_processed_gtfs_data() by simplifying code, adding more informative prints, and added calendar and calendar_dates as optional tables to supplement changes made in calendar file handling #70
  • refactor edge_impedance_by_route_type() by simplifying code, adding more informative prints, updates to accept new GTFS modes 11 and 12, new validation on input parameters, updates to docstring
  • refactor _interpolate_stop_times() section that uses _check_if_index_name_in_cols() for added clarity and fix edge case where index was removed before merge
  • throw an error when interpolater sees duplicate stop_sequence and unique_trip_id records to aid in debugging
  • modify informative prints on selected trips in _trip_schedule_selector() to be conditional on what input data is utilized
  • refactor create_transit_net() with:
    • minor formatting updates
    • docstring updates
    • print updates
    • simplification of repeated code
    • add prints to clarify when overwrite_existing_stop_times_int or use_existing_stop_times_int are used and dont allow overwrite_existing_stop_times_int and use_existing_stop_times_int to both be True for clarity
    • remove ValueError that could never be triggered

• general updates:

  • updates contributing guidelines
  • move create_transit_net() and headways() time range value check to its own shared function: _check_time_range_format() with unit tests
  • replace instances of '{}/{}'.format() -> os.path.join()
  • added and modified prints for functions that save and load HDF5 files in particular to simplify hdf5_to_df() key prints
  • address YAMLLoadWarning by replacing yaml.load(f) -> yaml.safe_load(f)
  • add new and updated unit tests for gtfs.network.py to expand coverage to 100%

Note: travis is failing in py27 and 35 for this PR due to issue with Pandana and Numpy

[smmaurer]

Looks like a great refactor! Code looks good, and tests pass on my Mac.

[smmaurer]

(We talked offline about the Travis failures in py27 and py35. The immediate problems are upstream, and will be fixed either in PR #87 or in a subsequent one moving the tests from Travis to GitHub Actions.)