Update installation and add FAQ. Handle warnings

[bocklund]

Fixes gh-126, fixes gh-128, fixes gh-134, and fixes gh-139

Regarding points in #126:

• Update installation to first suggest anaconda installation for all platforms. Also update the development mode, because conda-build is not well maintained for using development mode in environments. Suggested to use pip or setup.py in how to use conda develop? conda/conda-build#1992

This PR updates the installation guide with step by step instructions for all platforms, but emphasizing Anaconda. It is now more clear that you can use either distribution to install pycalphad.

The installation includes the caveats of the Ipopt installation situation for pip installation and develop mode, as well as a Anaconda alternative to develop mode.

• Add page describing parallelism with dask

This is in the newly added FAQ

• Estimate of computation time in the examples (roughly ~1 minute accuracy)

A brief statement of computation time and the implications for binplot/ternplot were added to the FAQ.

• (Possible) mention which warnings are normal to see, including convergence failures. Where should this go?

I did not handle the convergence failures (Ipopt should make them more rare). This commit explicitly adds code to filter out the NumPy warnings and fixes the dask import warning that users have been seeing and bumps the dask version accordingly. (Python 3.7 will make async a keyword (see https://bugs.python.org/issue30406). This means dask had to rename their async module to local to avoid having a module name be a keyword and users have been seeing warnings about the import. This commit updates the import from dask.async.get_sync to dask.local.get_sync and bumps the requirements to dask>=0.15.).

#139: Addressed simultaneously with #126

#128: added to FAQ

#134: Added a UsingCalculationResults example notebook (and generated docs) that goes through the calculate and equilibrium datasets and describes them fully. Also provides some brief examples of how one would select and mask using xarray.

[bocklund]

There were some minor code changes, but mostly docs here. I suggest checking out this branch locally and cd docs, then building the docs and inspecting them with something like sphinx-build . _build/html && cd _build/html && python3 -m http.server && cd ../..

[richardotis]

spelling

[richardotis]

• Al-Fe phase diagram picture on the index page is a broken link. (It's been that way a while.)
• Anaconda is a scientific Python distribution by Continuum Analytics. They call themselves Anaconda, Inc. now.
• This can be fixed in the FAQ needs a trailing colon.
• All the examples need new titles; shorter would be better. "Calculation" is in almost every one and doesn't help the user very much. (Almost every feature of pycalphad involves calculation.)
• Using Calculation Results: Headings should be calculate() Results and equilibrium() Results to be unambiguous
  -- Masking (mentioned at bottom) needs an example
  -- It would also be nice to mention .values for conversion to a NumPy array

[bocklund]

I think with the state of the solver, Al-Ni is a better looking example to show rather than Al-Fe. As far as the masking, the where is what that is: everything that does not match the condition is nan.

[richardotis]

Either phase diagram is fine. We just need to make sure the link on the front page is correct.

[bocklund]

TODO: Add notes on units in the FAQ

[bocklund]

This is ready for another look

[richardotis]

Merge when ready.