fix: refresh crons

[xavier-brochard]

Fix background event scheduling in send flow when UI is closed

Problem

• In some cases, the refreshSend background event was continuing to run even when the UI was closed/interface context no longer existed
• This caused unnecessary processing and potential errors when trying to update non-existent interfaces
• Events were being scheduled before successful completion, leading to continued execution even on failures

Changes

Background Event Handling:

• Added interface context existence checks in both refreshConfirmationEstimation and refreshSend handlers
• If no interface context is found (UI closed), the background events now exit early without scheduling the next run
• Moved background event scheduling to occur after successful interface updates rather than at the beginning
• Properly remove the interface id from the state after transaction success / failure

Result

Now refreshSend properly terminates when the UI is closed + on transaction success / failure, preventing unnecessary processing. Events only continue running when there's an active interface to update.

[sonarqubecloud]

Quality Gate failed

Failed conditions
0.0% Coverage on New Code (required ≥ 30%)

See analysis details on SonarQube Cloud

[darioAnongba]

Approved. There is something similar in the Bitcoin Snap where we gracefully exit when the interface is not found during the background tasks scheduling. I'm surprised you need to store the interface IDs though, but that's outside of the scope of this PR.

Also note that by exiting without throwing, MM will think the action succeeded and for example move the user to the activity tab. If you throw, then you'll see some errors appearing in sentry for no reason. I think we need to implement something on MM side so that we can return an error in the Snap.