Add upgrade test for legacy post-close monitor update persistence

[GideonBature]

LDK prior to 0.1 used u64::MAX (LEGACY_CLOSED_CHANNEL_UPDATE_ID) as the update_id for any ChannelMonitorUpdate generated after a channel had been closed. Since the change to stop using a constant monitor update_id after closure, we no longer generate these updates. However, both ChannelMonitor::update_monitor and MonitorUpdatingPersister still need to handle them so that nodes upgrading from pre-0.1 versions continue to work correctly.

This adds legacy_closed_channel_update, which synthesizes a pre-0.1 on-disk artifact: a ChannelForceClosed update with update_id == u64::MAX. The test exercises both directions of the legacy handling:

• Read path: The update is written directly to the KV store as a standalone update file named by the sentinel ID, matching what a pre-0.1 persister may have left behind. When the monitor is read back, it must discover and replay the update, leaving the monitor at update_id == u64::MAX.

• Write path: Passing the persister a sentinel-ID update must result in a full monitor write instead of a standalone update file. It must also purge stale update files through the legacy cleanup branch.

Closes #4104

[ldk-reviews-bot]

I've assigned @wpaulino as a reviewer!
I'll wait for their review and will help manage the review process.
Once they submit their review, I'll check if a second reviewer would be helpful.

[ldk-claude-review-bot]

I reviewed the entire diff and cross-checked the test against the implementation it exercises.

No issues found.

The test correctly exercises both legacy-handling paths:

• Read path: the standalone u64::MAX update file is discovered and replayed by maybe_read_channel_monitor_with_updates and applied via update_monitor, which sets latest_update_id = u64::MAX. Asserts match this behavior.
• Write path: passing a u64::MAX update to update_persisted_channel triggers a full monitor write (not a standalone update file) and the legacy cleanup branch cleanup_stale_updates_for_monitor_to(u64::MAX), which purges all update files. The list-empty assertion holds because TestStore::remove_internal ignores the lazy flag and removes synchronously.

Verified: field visibility (ChannelMonitorUpdate fields are pub/pub(crate), accessible in-crate), ChannelForceClosed { should_broadcast } variant shape, UpdateName round-trips u64::MAX correctly, and TestStore::new(false) (non-read-only) permits the writes/removes the test performs.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 86.97%. Comparing base (43ea836) to head (2963bd6).

Additional details and impacted files

@@           Coverage Diff           @@
##             main    #4755   +/-   ##
=======================================
  Coverage   86.96%   86.97%           
=======================================
  Files         161      161           
  Lines      111648   111715   +67     
  Branches   111648   111715   +67     
=======================================
+ Hits        97099    97160   +61     
- Misses      12045    12050    +5     
- Partials     2504     2505    +1     

Flag	Coverage Δ
fuzzing-fake-hashes	8.43% <ø> (ø)
fuzzing-real-hashes	32.41% <ø> (ø)
tests	86.30% <100.00%> (+<0.01%)	⬆️

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.