[JOSS review] Support reading/writing networks from different formats, integration with NetworkX/OSMnx

[martibosch]

The in_data argument of spaghetti.Network lists six possible input types but only describes four of them.
The notebooks of the tutorials use example data from PySAL in the form of a shapefile and iterables of libpysal.cg.Chain, nonetheless, the users might have their network data in other (more) common data formats (see, e.g., the formats supported by NetworkX). Would it be possible to instantiate spaghetti.Network objects from NetworkX graph objects so that spaghetti delegates the I/O part to NetworkX and focuses on the spatial analysis functionalities (see also #580). I know that there are other Python librares to deal with NetworkX, but since I suppose that spaghetti is mostly concerned with street networks, supporting NetworkX objects has the major advantage of supporting in turn OSMnx, which makes it very easy to stream street networks from OSM directly into Python objects.

Again, the developers might have good reasons (or other priorities) to not work with NetworkX objects, the point of this issue is to ensure that the matter has been considered.

openjournals/joss-reviews#2826

[jGaboardi]

Thank you for pointing this out, we updated the documentation to be more clear (see #592)

str, iterable (list, tuple, numpy.ndarray), libpysal.cg.Chain, geopandas.GeoDataFrame

We also added a wishlist/roadmap item to the issues (#593) to address reading these different formats in future. However, due to spaghetti being tightly integrated into the PySAL ecosystem, we believe that the majority of use cases are covered by the standard pysal input file formats such as shapefiles and GeoDataFrames.