Fix issue with gpkg reconstruction from merged diffs

[varmar05]

The issue:

In certain cases full gpkg file could not be reconstructed from diff checkpoints as those were incorrect.
Typically it happened when diff checkpoint started at version of basefile (e.g. v1) and during its construction from lower ranks starting diff was ignored leading to invalid checkpoint file.

Imagine situation like this when we wanted to create some higher rank checkpoint:

Checkpoint(rank=2, index=1, versions=v1-v16), 
[
Checkpoint(rank=1, index=1, versions=v1-v4), <-- was not created in recursion!
Checkpoint(rank=1, index=2, versions=v5-v8), 
Checkpoint(rank=1, index=3, versions=v9-v12), 	
Checkpoint(rank=0, index=13, versions=v13-v13), 
Checkpoint(rank=0, index=14, versions=v14-v14), 
Checkpoint(rank=0, index=15, versions=v15-v15) 
Checkpoint(rank=0, index=16, versions=v16-v16)
]

where first checkpoint in the list was originally ignored.

Fix:

This PR lessens the condition when checkpoint should be ignored only to situation if it spanned over basefile version.

[coveralls]

Pull Request Test Coverage Report for Build 23799172936

Details

• 22 of 22 (100.0%) changed or added relevant lines in 2 files are covered.
• 1 unchanged line in 1 file lost coverage.
• Overall coverage increased (+0.003%) to 93.241%

---

Files with Coverage Reduction	New Missed Lines	%
server/mergin/sync/models.py	1	93.92%

Totals
Change from base Build 23786864973:	0.003%
Covered Lines:	9133
Relevant Lines:	9795

---

💛 - Coveralls