Preparing v0.5 release

[smmaurer]

This PR prepares the v0.5 release, per issue #139.

Bug fixes

• Reinstates alternate names for aggregation types (issue Reinstate alternate names for aggregation types? #133)

Versioning and docs

• Updates version numbers in setup, init, readme, and docs
• Adds a changelog entry
• Updates the doc build process along the lines of other UDST repos
• Adds documentation for putting out a release (copied from other UDST repos)
• Updates the main README and revises the documentation -- the v0.5 docs are live now here
• Clarifies documentation of the decay parameter in network.aggregate(), per issue Exponential decay constant #128
• Revises the demo notebook

Dependencies

Pandana has accumulated some weighty dependencies that are not required for its core functionality of numerical calculation for simulations. In this release we're trying to slim things down.

• Drops the scikit-learn dependency by using an equivalent scipy function in network.get_node_ids()
• Makes matplotlib an optional install -- basemap was already optional because of deprecation issues, so this makes things consistent and also prints a helpful message when users run network.plot() with matplotlib or basemap missing
• Makes osmnet optional as well -- osmnet lets users load data directly from OSM, but it requires the whole geospatial python stack (fiona, shapely, etc) and often causes installation problems

Testing

• Updates the Travis script for Python 3.8
• Pins pytest-cov to an earlier version to resolve an incompatibility with coveralls, similar to this
• Changes a pep8 reference to pycodestyle
• Fixes a bug in setup.py that made it hard to run tests locally -- now you can run them with
  python setup.py test

Next steps

• Merge this to dev
• Release on PyPI
• Update the conda-forge feedstock
• Merge to master
• Tag release in GitHub

[lucky-verma]

When's the version 0.5 releasing?
I'm very eager to use some new methods in my project.

[sablanchard]

looks good!