Problem

Whenever we call a dokku subshell, the DOKKU_APP_NAME arg persists. As the official remote client uses --app functionality, the following will break a bit:

dokku couchdb:link service-name # run locally from a dokku app

Why does it break? The above call is translated to the following:

dokku --app APP couchdb:link service-name

Which then will do weird stuff if you call a subshell (as we do for docker-options and config set) within the couchdb plugin:

# this is called
dokku config:set APP KEY=VALUE

# but since we have `DOKKU_APP_NAME` set, the `dokku` binary turns that into this:
dokku config:set APP APP KEY=VALUE

Solution

Rather than sometimes calling the dokku binary and sometimes not, I figure we should move to calling the sourced functions directly so that we standardize on one way of doing things.

The alternative would be to always set DOKKU_APP_NAME to an empty string when calling a dokku sub-shell, which we'll totally forget to do.

Implementation

• removes the following CLI commands
  • dokku build
  • dokku receive
  • dokku release
  • dokku tar:build
  • dokku tar:build-locked
  • dokku git:build
  • dokku git:build-locked
• implements a central locking function for use in both git and tar plugins
• calls aforementioned locking function in receive-app trigger in given receiving plugin

Closes #2166