Refactor granularity kwarg in `protein.process_dataframe`

[ericmjl]

trafficstars

I think you're right about granularity - it's doing too much. The idea is to have control over:

1. What the nodes are in a residue graph (e.g. a-Carbon "CA", b-Carbon "CB")
2. Whether to use that atom position or compute residue centroids
3. Whether or not to build a residue-graph as above or an atom-graph

Originally posted by @a-r-j in https://github.com/a-r-j/graphein/issues/27#issuecomment-752674289

Refactoring out this little piece of the function will probably help with longer-term maintenance.