Add multi-query optimization to index

Currently JanusGraph supports multi-keys and multi-slice queries. However, not for index queries.
For example, let's assume we have a composite index on vertices for property `id`.
The below query will execute 5 sequential queries instead of doing either 5 parallel queries or a single combined query:
```
g.V().has("id", P.in(1,2,3,4,5))
```
I believe we can reuse similar technique we use for slice queries and apply this optimization for index queries as well.
As such, storage implementation will have a choice on how they want to execute such queries (i.e. in parallel, sequentially, or grouped query).

Start by reviewing the existing multi-key and multi-slice query handling, then trace how index queries execute values from P.in(1,2,3,4,5). Done means index queries can use a combined optimization while leaving storage implementations free to choose sequential, parallel, or grouped execution.

Written by the indexing model from the issue text.