Make Complex From Objects faster by creating a table and indexing

[thcrock]

If a feature's from object is complex enough to take even a few seconds, it can greatly increase the execution time if a lot of as-of-dates are present. This is because everything is parallelized over the as-of-date, so 1000 as-of-dates means the same query gets executed 1000 times, both for initial feature generation and the UNION ALL at the end ( https://github.com/dssg/triage/blob/master/src/triage/component/collate/spacetime.py#L214 ).

We can speed this up by preprocessing to create tables out of these and indexing them.