More Robust Statistics

[nmagerko]

The statistics we keep track of are currently generated on-demand and are then cached for a period of time. Ultimately, we should keep track of our statistics as they change, making it so that we'd (at worst) only ever have to compute our statistics once at the time of this enhancement's deployment.

Expectations for a good implementation might include:

• Rewriting the StatsService to write stats updates to the redis store and write-through to a stats database table
• Rewriting the StatsService to remove all joins and instead read from the redis store, pulling from the stats table only when necessary
• Deciding on new stats that we should be collecting

Note that we operate on multiple instances, so this implementation should be mindful of introducing race conditions into the stats aggregator.

[nmagerko]

This would also make it possible for us to show off how many applicants we're getting from each school in real-time. That would be pretty cool.