Add support for Fly Volume Restores

[davissp14]

This adds support for Fly-based volume restores.

How it works

The initial cue for kicking off a restore is the presence of the FLY_RESTORED_FROM environment variable. When this environment variable is present, we will push through the following process.

1. Check to see if the restore.lock file exists. The restore process will begin if the restore.lock file does not exist, or it exists but its value does not match the name of our app.
2. Reconfigure the pg_hba.conf file so we can bypass authentication locally.
3. Boot Postgres into standalone mode.
4. Create and update the flypgadmin, repmgr and postgres users and ensure their credentials match the current environment.
5. The repmgr database will be dropped so we can start a fresh cluster setup.
6. Revert the pg_hba.conf file back to its original configuration.
7. Shutdown the standalone Postgres
8. Create the restore.lock file and set its value to the name of our app. This will ensure the restore process isn't retried.
9. Begin the normal boot process.

Notes

• When restoring from a snapshot using flyctl, it's recommended that you restore into a single node cluster. Scale-out as needed once the restore completes.
• The credentials for the flypgadmin, repmgr and postgres will not be persisted across restores.