chore: Fly refactoring; fixes; extra LSN shipping coverage

[pcnc]

• Refactors LSN checkpoint creation and handling
  • Introduces handler that regularly ships the checkpoint file - used as a backup for cases such as volume corruption
    • Currently at 10 minute intervals, if the latest LSN has updated in the meantime
  • Ships the checkpoint after supervisord exits, as a backup to when signal handling goes awry
• Some fixes around perms and order of operations when unpacking the init payload

[darora]

Testinfra AIO tests were passing in the past I think?

[bmpandrade]

Still failing test-infra-aio with:

E           requests.exceptions.ConnectionError: HTTPConnectionPool(host='localhost', port=8000): Max retries exceeded with url: /rest/v1/buckets (Caused by NewConnectionError('<urllib3.connection.HTTPConnection object at 0xffffbc858f70>: Failed to establish a new connection: [Errno 111] Connection refused'))

Looks like services are taking longer to start and it fails.

@pcnc , will look into the test suite and see if there is a hold we can do until services are up and running, or if the service itself failed to start.

[bmpandrade]

In general it looks good, but I got a bit confused about the db-only as referred on the other code comments.

I thought, that only on special cases db-only was used, and the fly extensions was not one of them. Wouldn't this make us update the worker too, when deploying to Fly to treat it as a db-only, to ensure that LSN checkpoints were being properly uploaded?