Token filtering gets very slow

[wontonst]

I have a query that does

Token.objects.filter(user_id=blah, expires_at_gte=current_time)

This query becomes very slow at scale. I have some users that log in multiple times a minute which over time accumulates over a million rows.

This issue is a potential fix

#60

but as mentioned the token history is lost.

My proposal is to add a composite index so filtering is faster.