fix: Run database migrations before starting anything else

[joshlarson]

This allows us to put migrations and code that relies on those migrations into the same commits.

If the code relies on changes made in a migration, and the migration runs after the code starts, then there's a potential race condition where that code runs before the migrations have run, which would cause an error. This eliminates that possibility.

[github-actions]

Coverage of commit 329d71e

Summary coverage rate:
  lines......: 93.1% (3319 of 3566 lines)
  functions..: 72.6% (1357 of 1870 functions)
  branches...: no data found

Files changed coverage rate:
                                                                  |Lines       |Functions  |Branches    
  Filename                                                        |Rate     Num|Rate    Num|Rate     Num
  ======================================================================================================
  lib/skate/application.ex                                        |94.1%     17|75.0%     4|    -      0

Download coverage report

[firestack]

I can see that maybe this needs to be done before the Endpoint is started, but does it need to come before the PubSub?

Also, when would code in the Endpoint conflict with the Migrations? I assume this would never happen in a deployment due to our health checks?

[joshlarson]

Basically I was thinking that code in the Endpoint would refer to some column in the database that's introduced in a migration. That'll cause an error in the

Do the healthchecks fail if there are any un-run migrations?

My reason for this was remembering that a while ago, when I added a migration, I was asked to pull the migration out to a pre-PR that got deployed before the main PR was merged. Maybe that wasn't necessary though?