Proactively cleanup garbage files during write abort

Related to https://github.com/apache/fluss/pull/1441#discussion_r2280929949

When aborting a write, we could proactively cleanup garbage files.

It is debatable if the format or the engine should do this, for example Iceberg does both:

Task writer abort: https://github.com/apache/iceberg/blob/main/core/src/main/java/org/apache/iceberg/io/BaseTaskWriter.java#L93

Spark cleanup: https://github.com/apache/iceberg/blob/main/spark/v3.4/spark/src/main/java/org/apache/iceberg/spark/source/SparkCleanupUtil.java#L56

but currently because Lance uses object_store, and it does not really have a binding for Java, either Lance needs to expose the cleanup hook for engine to invoke, or engine has to configure another file system library to do cleanup, both not ideal.

My take is that we probably also need both:

For process that directly call Lance library to perform write, e.g. merge-insert, it makes sense that the merge-insert job does its own cleanup in case of any failure.

But if engine has any custom logic and want to control the cleanup process, they should have a way to do so. For that, I think the work in OpenDAL https://github.com/apache/opendal/issues/6456 is very useful, so we can make OpenDAL basically the file system SDK for engines to interact with Lance.

Start with the linked pull request discussion and OpenDAL issue 6456, then compare the cited Iceberg task-writer and Spark cleanup paths. Define how aborted-write cleanup should serve direct Lance callers such as merge-insert while allowing engines to control cleanup; done means the ownership and integration approach are agreed and implemented.