[hf] cleanup migration flow tooling

context

we just migrated a subset of channels onto a fresh instance. in the process, we noticed some sharp edges in the tooling for migrating from one box to another. This MR attempts to sand and polish those edges. :)

changes

• cleanup docker install in ansible.provision
  • ensure that pip role uses python3-pip
  • restart docker at end of provisioning to try to prevent EAI_AGAIN errors during deploy step
• set NODE_ENV to production during db restore
  • context: repeated runs of restore were failing because the /bin/restore script in isolation creates databases with NODE_ENV set to development. (it only works on the first run b/c the deploy playbook created the correct dbs)
  • to allow us to drop db's and recreate, we enable restore to target the correct environment on its own (so that deploy need not be re-run to allow restore to succeed)
• update inventory.tmpl to include new boostit instance ip