Addresses #6634

There's a possibility that users start a stopped cluster manually through the cloud provider's UI or API.

When this happens, and we try to refresh the cluster status, the following happens:

1. We see that all instances are up from the cloud API
2. So we proceed to check ray status, but this will fail with SSH connection timeout, because the manually started instance could get a new IP address, but we only know the old one

Other than having a different IP address (which could be resolved simply by re-fetching the latest data from the cloud API), the other problem with manually started clusters is that it doesn't have the SkyPilot runtime installed (will error with Ray cluster is not found ...). So the only way to recover from this is to do sky start again.

This might not be intuitive though for users, so this PR adds a log message to help them recover from this state.

IP address change could also happen if a user manually allocates a static public IP address after sky launch. In this case, we also won't know of the IP address change, until the user does sky start again.

Tested (run the relevant ones):

• Code formatting: install pre-commit (auto-check on commit) or bash format.sh
• Any manual or new tests for this PR (please specify below)
  • Added test_aws_manual_restart_recovery
• All smoke tests: /smoke-test (CI) or pytest tests/test_smoke.py (local)
• Relevant individual tests: /smoke-test -k test_name (CI) or pytest tests/test_smoke.py::test_name (local)
• Backward compatibility: /quicktest-core (CI) or pytest tests/smoke_tests/test_backward_compat.py (local)