Ensure the operator is connected to the prover before sending work

[jules]

This avoids an infinite loop in which a worker continually asks for work but cant get it back from the pool as it is not properly connected.

[ljedrz]

Other than the fact that I agree that we should instead have something like https://github.com/AleoHQ/snarkOS/issues/1554 in place, this workaround should be adjusted as follows:

1. the first few early returns in Peers::update should send Err(<some meaningful error>) to the Sender in connection_result, so we can actually know if we weren't already connected or if there's some other connection attempt error
2. instead of ignoring the result of the corresponding Receiver (let _ = handler.await;), we should check whether we succeeded at connecting, were already connected or if there was an error and potentially act differently if ultimately we're not connected after all

[ljedrz]

The CI jobs should succeed once https://github.com/AleoHQ/snarkOS/pull/1625 is merged and this PR is rebased on top of it.

[aleo-svc-exp]

Admin: Please respond with "build this" in order to build this PR.