Allow ftl config/secret to run without the controller

[deniseli]

• Use controller if it’s running
• If you’re in production, you don’t want to use the local config/secret setting logic; you want it to fail immediately
• Client needs to know if a local vs prod controller is running. To do this, look at the endpoint value (e.g. ftl deploy --endpoint). If you give it a production endpoint, then it’ll talk to a production controller. We could look at the endpoint value - if it’s localhost and not running, then do local stuff. Let’s do that for now.
  • A real prod example: ftl --log-level=trace --endpoint=https://ftl.tbddev.org/ ps
  • We should rethink this approach (relying on --endpoint) at some point, but this is fine for now and we want to get the fix out quickly to avoid hurting the UX for PFI

tl;dr:

1. Abstract the config/secret part of the service behind an interface
2. If controller is not present AND endpoint is on localhost / 127.0.0.1 ... might need to use netip.Addr.IsLoopback() to be sure, inject an impl of that interface that operates locally, otherwise inject the existing gRPC client