Retry NM-owned retryable suspend failures before releasing network teardown

[Staphylococcus]

Summary

• give NetworkManager-owned retryable suspend failures the same single follow-up opportunity as joined failures
• record a completed cycle and ownership marker when that follow-up succeeds
• release teardown after one failed follow-up without introducing a retry loop

Test-driven evidence

The owner-first regression initially failed with:

left:  Some(RetryableTransportFailure)
right: Some(Completed)

After the implementation change, both the recovery case and the bounded second-failure case pass.

Validation

• cargo fmt --all --check
• cargo clippy -p lg-buddy --all-targets -- -D warnings
• cargo test -p lg-buddy --lib (737 passed, 1 ignored)
• cargo test -p lg-buddy (124 Cucumber scenarios / 1,431 steps plus all integration tests)
• git diff --check

Closes #121