Optimize event study code

[jonathandroth]

Calculation of the variances and other such objects could be done only once across event-times, leading to computational improvements for event studies

[mcaceresb]

@jonathandroth The Stata version is currently about 10x faster but the code-base is significantly different.

Before doing a full re-write though I can try optimize the code around the margins (e.g. avoid computing the same thing multiple times, etc). LMK if I should dive into the refactor or try optimizing first.