Add AFLOW functionality to propnet

[clegaspi]

• Adds "adapter", à la MPRester, which can produce propnet materials from AFLOW IDs. Can also produce raw data from AFLOW that has been converted to native Python types. Special field names which match propnet symbol names are used for AFLOW properties based on external files, such as structure (returns pymatgen Structure object) and elastic_tensor_voigt and compliance_tensor_voigt (returns list of lists).
• Adds database ingester (propnet.dbtools.aflow.AflowIngester) as a way to download the AFLOW database, limiting the number of API requests by breaking up data requests into smaller chunks. Based on Builder, so stores data in a MongoDB in its native format (strings), except external JSON files are stored as dictionaries.
• Adds the ability to retrieve a material by AFLOW ID on the web app and calculate the propnet graph.
• Fixes to Model and Graph for bugs revealed during AFLOW adaptation.
• Adds a few new models and symbols (Thanks @dmrdjenovich)
• Updated input set generation algorithm in Graph to reduce some of the slowness in the graph algorithm which was exposed by this adaption. However, ultimately there is an issue with combinatorics that needs to be addressed in another major update. See Add aggregation pipeline in Graph evaluation #238 for details.

[coveralls]

Coverage increased (+1.1%) to 85.93% when pulling 00080fb on aflow_adapter into 759b193 on master.