Improve tests for unaligned checkpoint sync

[michaelsproul]

Description

We had a latent bug in checkpoint sync from unaligned blocks (skipped slots at epoch boundaries), fixed here:

• Reimport the checkpoint sync block #8417 (comment)

I don't know why the existing tests did not catch this:

• weak_subjectivity_sync_unaligned_advanced_checkpoint
• weak_subjectivity_sync_unaligned_unadvanced_checkpoint

To resolve this issue, we can try reverting the fix made in the linked PR, and writing a test that fails sans fix.

[Andrurachi]

Hey! I’d like to take a shot at this if it’s open to outside contributors.
I can dig into the tests and make sure the regression is properly caught after locally reverting the fix.

Mind if I pick this up?

[michaelsproul]

@Andrurachi yeah for sure, thanks!

[michaelsproul]

Just to keep you posted, we changed the fix slightly to work around some awkward invariants that LH maintains:

• Reimport the checkpoint sync block #8417 (comment)