Add SQL query counts to tasks and API

[dcramer]

This puts some basics in place to let us do production based perf assertions.

The change itself adds duplicate queries (>= 3) and total queries (>= 25) to tasks and API endpoints.

@getsentry/api @getsentry/infrastructure

[dcramer]

probably should refactor this but its not a huge deal

[dcramer]

One odd behavior this creates is our integration tests effectively nest context managers, so i.e. the store_event task ends up suggesting it executes a LOT of queries, but really its just the sum of all child tasks.

[codecov-io]

Current coverage is 82.40%

Merging #2370 into master will increase coverage by +0.08% as of 7ced1e2

@@            master   #2370   diff @@
======================================
  Files          932     934     +2
  Stmts        36641   36718    +77
  Branches         0       0       
  Methods          0       0       
======================================
+ Hit          30163   30257    +94
  Partial          0       0       
+ Missed        6478    6461    -17

Review entire Coverage Diff as of 7ced1e2

Powered by Codecov. Updated on successful CI builds.

[dcramer]

@mitsuhiko @tkaemming if this seems sane to one of you guys i'll go ahead and push it out

[tkaemming]

Does this really need a cryptographic hash rather than hash?

[dcramer]

i always forget that python maps strings to the same

[tkaemming]

Generally +1. (I didn't try and run this or give it a super thorough review though.)

I could see an argument for keeping track of and logging the duplicate query statements as well, but that's easy enough to do later.

[tkaemming]

Also putting this in sentry.perfutils rather than sentry.utils.performance or something seems to give it more priority than I'd expect (but that's just my opinion.)

[dcramer]

I may just put it in sentry.debug idk

[dcramer]

Going to get this in today

[dcramer]

I'm thinking we need a way to whitelist things, but we can sort that out later. It's likely this will pollute Sentry a bit right now, so we'll run it for a day or two and then decide how we want to deal with it.