lightning: fix several disk-quota-related bugs

[kennytm]

What problem does this PR solve?

1. Fix Lightning Disk Quota is not fully respected after restart due to non-deterministic engine selection #812
2. Fix the bug where the checkpoints in local back-end are not advanced even after disk quota force-flushed all engines.

What is changed and how it works?

1. For Lightning Disk Quota is not fully respected after restart due to non-deterministic engine selection #812, it is caused by iterating cp.Engines in restoreEngines(), which is a hash map. We read the map into a sorted array and iterate that instead, so engines are always restored in deterministic order.
2. For the checkpoint issue, we now update the checkpoint based on the Local Writer's synchronization status instead of the disk quota status (the deliverLoop cannot see diskQuotaStateImporting during diskQuotaLock.RLock()). This assumes Pebble's Ingest() will persist all content to disk.

Check List

Tests

• Integration test (TODO)
• Manual test (add detailed scripts or steps below)
  • Testing its effect in the 1T GA test.

Code changes

Side effects

Related changes

• Need to cherry-pick to the release branch

Release Note

• No release note.

[kennytm]

Separated into #761, #845, #846, #847, closing.