fix: v2 deadlock when there are multiple data files

[westonpace]

We read batches from data files in a round-robin fashion. With the old priority mechanism all data in file 1 would be scheduled before file 2. This means we'd deadlock if file 1 was bigger than the I/O buffer size.

Now, each fragment has a unique priority, but the priority within the fragment is the same for all data files.

This should avoid deadlock as long as the I/O buffer size is larger than any given batch (may want to issue warning or fail if we can detect when the batch size is too large).

[codecov-commenter]

Codecov Report

Attention: Patch coverage is 70.00000% with 3 lines in your changes missing coverage. Please review.

Project coverage is 77.80%. Comparing base (5606b17) to head (e1bc463).
Report is 1 commits behind head on main.

Files with missing lines	Patch %	Lines
rust/lance-io/src/scheduler.rs	0.00%	3 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #2923      +/-   ##
==========================================
+ Coverage   77.78%   77.80%   +0.01%     
==========================================
  Files         231      231              
  Lines       70199    70188      -11     
  Branches    70199    70188      -11     
==========================================
+ Hits        54607    54609       +2     
- Misses      12688    12694       +6     
+ Partials     2904     2885      -19     

Flag	Coverage Δ
unittests	77.80% <70.00%> (+0.01%)	⬆️

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

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