OSMnx tweak, address masking issue/warning

This release just addresses 2 reported issues:

• OSMnx API had changed so the plotter was broken
• There was a warning being emitted when using a mask operation in one of the graph analysis workflows

Pandas support, new attributes and configurability

• Housekeeping upgrades for pandas, numpy, scipy
• New attributes on nodes (modes at node as list)
• Allow custom wait time as a parameter
• Logger on coalesce function

Wait time estimation and handling improvements

• Reduce processed trip selection to only those within service times parameters (85daf75)
• Wait time estimation ignores overweighting of zero values (e9b038f)
• Wait time estimation can account for service time start and end parameters (a9ef0bd)
• Safer direction id handling (0c5073d)
• More careful testing around single stop/arrival stop-schedule pairs (fbd32f1)

Note: Also included were the two changes included in v0.6.1 which failed to get a release on Github. Updates pertinent to that version are the following:

• Auto detection of transfer points when ingesting custom transit line transitJSON
• Pinning partridge dependency to >= 1.0.0 with new Feed class

graph-tool export support and improved transitJSON validation

• Update partridge dependency
• Improved Docker container creation (support for graph-tool)
• graph-tool export feature
• Class-based transitJSON validation
• Consolidate OSMnx dependency
• Resolve CRS reset bug
• Significant increase in documentation for library methods

Support single trip timeframes

• Handle when only on trip exists within the target time frame
• Avoid errors related to no valid wait times being able to be parsed (due to single trip)

Significant performance optimizations

• Reduce stop times calculations within route processing through timeframe trimming
• Remove append operations on pandas DataFrames inside RouteProcessor
• Overall 2/3 reduction in run time, on average, to process a new GTFS feed
• Resolve direction ids being NaN only some of the time by dealing with direction id on a per-route level
• Parallelization of intermediary stop times imputation

Save and read parsed feeds, additional multiprocessing

Primary enhancements:

• Multiprocessing support for interpolating stop times
• Read and write parsed network graphs to disk

Improve import experience

• Resolve annoying behavior on import with matplotlib being called and a figure plotted immediately on import

Custom synthetic stops, general refactoring

• Custom stop coordinates for synthetic routes
• Drop duplicate edges during simplify operation
• Updates to Makefile scripts
• Assorted refactoring

DiGraph conversion, simplification, MP optimizations

• Refactor map operation for multiprocessing pool
• MultiDiGraph to DiGraph converter
• Simplify graph method
• Refactor of coalesce to deal with lost edges problem