[fix] Skip Auto-Observe on Failed Broadcast

This is a small change to only auto-observe signed transactions that have been successfully broadcast. We initially auto-observed regardless of failure, since broadcast can error when the transaction made it to the mempool, but we have observed cases where it seems a subsequent signing uses the same UTXOs and invalidates the previous transaction. Since we do not re-sign the same outbound with a new transaction containing different UTXOs (to prevent double spend), this results in the failed transaction continuing to retry broadcast with an invalid transaction until it is rescheduled to a new vault. This change should reduce excess auto observations so the outbound will be rescheduled to a new vault sooner.