Node Caching Feature

[jbryan]

Since there is no concept of "query" and certainly no query optimization, and since operations that require accessing a large number of nodes can be costly, it would be nice to be able to have some form of node and/or relationship caching such that repeatedly getting the same node can be a low cost operation. This may be an in-python solution, or it may be a memcache solution.

The exact details about when and how to invalidate the cache should be discussed here as well motivating use cases and benchmarks.