Add local caching

[scott-ainsworth]

I suggest that memgator be enhanced to cache retrieved timemaps. Reasoning:

• Repoducability: In the academic environment, the ability to repeat an experiment is important. Implementing a cache will allow memgator to send the exact same response even if archive content changes.
• Analysis: Having a copy of everything memgator consumed, produced, or both, enables analysis of the data used in an experiment.
• Debugging: It is also important when testing and/or debugging code. If the resulting combined timemap "doesn't look right", the cache provides an easy way to review the original timemaps.
• Avoiding duplicative effort: If memgator does not implement a cache, then users of memgator who require a cache will need to implement it. Since memgator already touches every timemap, having it implement a cache is efficient.
• Speed: Finally, using standard HTTP cache semantics, a local cache can speed up responses.

Note: I have a cache design that allows all of the above. It is written in Python, but should be easily ported to Go.

[machawk1]

:+1: but not as the default functionality. I enjoy knowing that each query supplies fresh data. A couple of different caching policies should be available to the user, e.g., neverRefreshOnceGenerated, refreshEveryIRequests, refreshWithATimeBasis, alwaysRefresh (default), etc.

@galsondor, can you provide your Python code or a local (on ws-dl-xx or otherwise) where I can find it? I have been looking for a reason to get my hands dirtier in the Memgator code and porting your implementation might be a good practical means to accomplish this.