Pry run and reload

[krainboltgreene]

• Integrate Pry.reload
• Document new commands
• Stop migrating on every request
• Replace the usage of db:setup, which doesn't exist, with db:migrate
• Split up CircleCI tasks so we get a status indicator for everything
• Building docker images so we know when we've fucked up the build
• Putting CircleCI into a workflow so we get one setup done before all the real work
• Made all bin/* commands also correctly report exit codes, that way CircleCI knows when something failed
• Created bin/local-each which runs a command for each service
• Splitting up postgres setup into two sections, one that cares about failure and one that doesn't
• Renamed commands that targeted all services with the suffix -all
• Renamed the _*/ directories to a unified .*/ like rubocop and circleci, except _base which made sense to rename as base.
• Moved the sorbet-runtime gem to no-group so it's used in production