Move docket view count out of docket model

[mlissner]

Right now we tally a count to the docket table every time somebody views a docket. I'm sorry to say this is a bad idea. Looking up, locking, and then incrementing a row on a table of this size every time somebody loads the page is difficult. In fact, I've noticed that this query is our worst query according to AWS:

(You can't really tell, but it's the first one you see there.)

I'm not sure the best fix for this:

• We could put this in redis.
• We could use our stats app for it.
• We could develop something new.
• We could use some kind of external API like Plausible or whatever (this seems really slow).
• We could use a special stats database for this (Clickhouse?)

We should do something though, I'm afraid....

[mlissner]

By the way, the #917 will also fix this problem when we do it (at the cost of losing the accuracy of this stat).