use Miniconda3

[jGaboardi]

Addressing #62 (see also pysal/splot#48)

Since there is no more support for Python2, is there a specific reason to perform TravisCI environment builds with Miniconda2? This PR addresses that issue and includes the new conda feature:

- conda config --set channel_priority strict.

[coveralls]

Coverage remained the same at 90.945% when pulling d621334 on jGaboardi:miniconda3 into 81a3bb5 on darribas:master.

[jGaboardi]

Python 3.5 builds are failing seemingly due to a dependency issue with rasterio.

@ocefpaf Might this current problem be related to the pillow/jpeg issue from several years back?

[jorisvandenbossche]

I suppose this is because conda-forge does not provide recent package any more for python 3.5 (there is only rasterio 0.36 available for python 3.5, no >= 1.0. Not sure why it wouldn't want to install this older version though).

So we could either drop 3.5 (related to https://github.com/darribas/contextily/issues/54), or either use defaults for 3.5.

[jorisvandenbossche]

@darribas although this is still failing, I think it is good to merge this. We want this update to miniconda3 anyway.

[ocefpaf]

@ocefpaf Might this current problem be related to the pillow/jpeg issue from several years back?

Nope. Like @jorisvandenbossche mentioned above the problem here is the dropped support for Python 3.5. I recommend to either test it on 3.6/3.7 only or move the testing to pip. In conda-forge we don't build all version of Pythons due to CIs constraints.

[darribas]

OK, merging then and thinking maybe we could drop 3.5? See poll over at #56

[jorisvandenbossche]

@darribas BTW, just a random side note (and feel free to ignore :-)): for PRs like this I think it can be beneficial to squash on merging (the green button of github can do that for you), as it are a bunch of back and forth commits that are not that informative (I personally regularly look at the git history to see what / when / why things changed).

[darribas]

Thanks @jorisvandenbossche. It's more my "gitnorance" than any other thing... I'll try to comply in the future as I agree, cleaner and more readable commits are a great thing!