Allow repartitioning on files with ranges

[Samyak2]

Which issue does this PR close?

• Fixes ExecutionPlan::repartitioned doesn't work when file groups have ranges #18940

Rationale for this change

When any partitioned file had ranges selected (even if the range contained the whole file), repartitioning was being skipped.

There was no good reason to disallow this.

What changes are included in this PR?

• I have now allowed this and change the implementation to respect file ranges.
• Also did a very small refactor to make the code easier to read: replace state tuple and state.0/state.1 with proper variable names using destructing

Are these changes tested?

Added a couple of unit tests. Let me know if more are needed.

Are there any user-facing changes?

[Samyak2]

I rebased on main. Not sure why cargo fmt is failing on a file that I did not touch!

 + cargo fmt --all -- --check
Diff in /__w/datafusion/datafusion/datafusion/execution/src/config.rs:114:

 /// A type map for storing extensions.
-/// 
+///
 /// Extensions are indexed by their type `T`. If multiple values of the same type are provided, only the last one
 /// will be kept.
-/// 
+///
 /// Extensions are opaque objects that are unknown to DataFusion itself but can be downcast by optimizer rules,
 /// execution plans, or other components that have access to the session config.
 /// They provide a flexible way to attach extra data or behavior to the session config.

[Jefffrey]

I rebased on main. Not sure why cargo fmt is failing on a file that I did not touch!

 + cargo fmt --all -- --check
Diff in /__w/datafusion/datafusion/datafusion/execution/src/config.rs:114:

 /// A type map for storing extensions.
-/// 
+///
 /// Extensions are indexed by their type `T`. If multiple values of the same type are provided, only the last one
 /// will be kept.
-/// 
+///
 /// Extensions are opaque objects that are unknown to DataFusion itself but can be downcast by optimizer rules,
 /// execution plans, or other components that have access to the session config.
 /// They provide a flexible way to attach extra data or behavior to the session config.

Unrelated error, fixed on main now so I've merged up from main

[Samyak2]

Thank you! Do I need to tag more people to review this PR? or let it be for a few days?

[Jefffrey]

I plan to review this PR soon (hopefully today or tomorrow)

[Jefffrey]

The changes look good to me. Only thing to comment on is perhaps adjust the test cases so the ranges don't span the entire file; for example, pfile("a", 123).with_range(1, 122) just as a quick sanity check that we respect ranges that don't essentially alias an entire file.

[Samyak2]

Thanks for pointing that out @Jefffrey! There was a bug in the order-preserved case that lead to some partitioned files being empty when the start of the range was non-zero. I have fixed the bug and added tests in the latest commit.

[Jefffrey]

I'll leave this PR open for a few days before merging in case anyone else is interested in reviewing it