fix: ensure node state persisted before shutdown

[jvsena42]

Fixes #739

This PR prevents channel state divergence during node shutdown by ensuring state is fully persisted before the service is destroyed.

Description

When the app was stopped while a 0-conf channel had uncommitted state updates, the client (LDK) could end up with a different commitment height than the LSP. On reconnect, the LSP detected this mismatch as "possible data loss" and force-closed the channel.

This PR adds two mitigations:

1. Final sync before shutdown - Calls syncWallets() before stopping the node to ensure the latest channel state is persisted to VSS
2. Blocking service shutdown - Uses runBlocking in onDestroy() to wait for the node to fully stop before the service is destroyed, with a 5-second timeout to avoid ANR

Preview

CJIT.webm

multiple-transactions-and-poor-signal.webm

QA Notes

1. Test graceful shutdown

1. Open the app and create or use an existing Lightning channel
2. Stop the app using the notification "Stop" button
3. Check logcat for:
   • Performing final sync before shutdown…
   • Final sync completed
   • onDestroy started
   • onDestroy completed
4. Verify the node stops without errors

2. Test restart after stop

1. After stopping the app via notification, reopen it
2. Verify the Lightning channel is still operational
3. Verify no force-close occurred

3. Regression

• Send and receive Lightning payments normally
• Verify wallet balance updates correctly

[jvsena42]

DIdn't find problems in my tests, but this has a timing very difficult to reproduce

[claude]

Code review

No issues found. Checked for bugs and CLAUDE.md compliance.

[ovitrif]

nit: Probably still good to use non-main thread

Look at runBlocking(dispatcher) option