Compatibility with Pandas v1.0

[msoltadeo]

Make updates for new version of Pandas:

• Use .reindex() instead of .loc when there are missing values
• change .as_matrix() for .values

[coveralls]

Coverage decreased (-0.1%) to 94.196% when pulling effd3a8 on update_pandas into 79f815a on master.

[smmaurer]

Some more context:

What prompted this PR is that certain syntax was removed in Pandas v1.0, causing parts of urbansim to stop working. (It had been deprecated for a while.)

Travis

In order to get the Travis tests to pass, I went through and fixed some other problems as well:

• Fixed a DataFrame comparison in a test (61b1d50) -- Pandas changed how DataFrame columns are ordered in one of the past several releases.
• Replaced an instance of .ix[] (2954eb2); I think this syntax was removed in Pandas v1.0 too.
• Made minor updates to the Travis script so that the environment can continue installing properly (ipython-notebook -> jupyter, etc).

I had to pin pytest at < 4.0 because of a problem with how some of the unit test fixtures take parameters that I couldn't figure out how to solve. This doesn't affect the library itself, just the procedure for running tests.

AppVeyor

I made similar environment updates for AppVeyor. Those tests run now (no longer blocked by package installation), but fail with a couple of string formatting errors that I'm not sure how to solve but don't seem important.

@janowicz @sablanchard, do you think it's ok if we go ahead and merge this, and open an issue to deal with AppVeyor later?

[smmaurer]

Managed to get everything working in AppVeyor; all the tests now pass.

Plan is to merge this into a dev branch, then start another PR with everything necessary for releasing a point update (versioning, docs, etc).