Add FileIO.list_prefix

[rambleraptor]

Rationale for this change

We've had a lot of interest in #3361. I'm attempting to split that into some smaller PRs to make it easier for everyone.

This PR adds a list_prefix as a method on a SupportsPrefixOperations mixin. Thanks to @Fokko and @kevinjqliu for helping to point out the history around FileIO.

Are these changes tested?

Includes unit tests.

Are there any user-facing changes?

AI Disclosure: Claude did help here.

[Fokko]

Thanks for bringing this up, this is a very sensitive topic 😅 Iceberg is very opinionated about targeting object stores, rather than filesystems. In object stores, list operations (list-prefix) are both slow and expensive because they are paged responses. Therefore, we don't want to make them first-class citizens in the FileIO. The list operation is also missing from the reference implementation.

We could follow the same path there, where we have something like SupportPrefixOperation.

I can see PyIceberg being used to clean up small, low-volume orphan files. Ideally, I think you want to use something storage-specific, like storage-inventory.

[rambleraptor]

@Fokko thanks so much for bringing this up! This makes absolute sense and explains why Java has the SupportPrefixOperation. I agree that - at large scale - using an inventory product can help you identify orphaned files, but it's really tied to your underlying storage provider. I think there's still a place for PyIceberg to give you a quick CLI to perform some basic maintenance tasks.