TST: add regression test for combine_first with duplicate index (GH#66009)

[AdvenRanises]

Adds a regression test for combine_first behavior with duplicate indices.

This ensures current positional alignment behavior remains stable and guards against future changes that might attempt to enforce uniqueness or alter duplicate handling semantics.

Closes #66009 (clarification via test, not behavioral change).

[krishgarg344]

Hey, tried this myself. If I make s2 int instead of float (with the same duplicate-index setup), combine_first raises ValueError: cannot reindex on an axis with duplicate labels. That comes from a check in Index.reindex that's tagged GH#42568 in the source.

So it looks like pandas has already decided that duplicate indices should raise here. The float/float case seems to skip that check entirely because it takes a different shortcut earlier in the function.

Feels like that's the actual bug, rather than something to lock in with a test. I might be missing some context, but wanted to flag it before this merges and closes #66009.

Here's the error when s2 is made int:

   4251 elif not self.is_unique:
   4252     # GH#42568
-> 4253     raise ValueError("cannot reindex on an axis with duplicate labels")
   4254 else:
   4255     indexer, _ = self.get_indexer_non_unique(target)
ValueError: cannot reindex on an axis with duplicate labels

[jbrockmendel]

why is a test locking down this behavior the right move? Doesn't the link issue treat it as a bug?