[ENHANCEMENT] Further improvements to graph indexing code logic

[mtfishman]

Here are some potential followups to #115:

• Vertex wrapper (like Vertices but indicating it is a single vertex).
• to_vertex function as an interface for canonicalizing single vertex inputs, such as converting QuotientVertexVertex to a plain vertex (which is currently handled by to_graph_index).
• get_graph_index could call out to new interface functions get_graph_vertex or get_graph_edge, and maybe call to_vertex or to_edge for vertex/edge canonicalization (see the point above proposing to_vertex).
• get_graph_indices could call out to new interface functions get_graph_vertices or get_graph_edges, which could call to_vertices or to_edges for vertex/edge collection canonicalization (such as wrapping QuotientVertexVertices in a QuotientVertexSlice).
• Reevaluate parent_graph_indices, maybe find a better name or see if it can be removed.
• Reevaluate graph_index_list_to_graph_index name/interface. Maybe it could be called flatten_graph_index_list?
• Define isassigned(g::AbstractNamedGraph, I) = false (and write it in terms of associated interface functions is_graph_index_assigned, etc. which would get overloaded by DataGraphs).
• Define and ispartitioned(g) trait function and use it in departition/unpartition, see the discussion here: PartitionedGraphs extension for DataGraphs and interface overhaul.  DataGraphs.jl#55 (comment).
• Reassess what is output by add/rem_vertex!, add/rem_edge!, since it is not consistent with the output of those functions in Graphs.jl. Maybe we want to introduce other stricter interfaces like insert_vertex!, delete_vertex!, set_vertex!, unset_vertex!, etc. (modeled after the Dictionaries.jl interface).