fix errors and automate manual aspects of provision and deploy process

In the process of setting up a new instance of clientcomm I ran into a few things that could be fixed, documented, improved, or automated:

[ ] Finalize and document the process for managing multiple deploys with terraform state environments
[ ] Could the deploy script automate the database migration and bower install detailed in the Initializing the app section of the deploy README?
[ ] This commented-out line in clientcomm.tf gets written into the clientcomm.conf file on the deployed servers, which errors if you try to export the file to env variables
[ ] Document that a new hosted zone must be created for the new deploy's URL (i.e. sanfrancisco.clientcomm.org, and an NS record be created in the clientcomm.org hosted zone with the new zone's name servers. (This is because there's a wildcard A record in the clientcomm.org hosted zone that intercepts requests for subdomains unless they're explicitly noted)
[ ] Once the mailgun record sets are created, you'll have to verify the domain setup on the mailgun site. Also, the mailgun SMTP password (TF_VAR_mailgun_smtp_password) is a value that can be retrieved from mailgun and is expected in the environment variables. These details should be added to the deploy README.
[x] The initialize deploy script isn't working on the new servers; it hangs when trying to access the database. Figure out why and fix it.

slco-2016 / clientcomm