Add EmptyMatchTreatment support for SMB reads (#5759)

[sushmullur]

Summary

Adds support for Beam's EmptyMatchTreatment parameter in SMB read
operations, allowing pipelines to handle empty directories gracefully
instead of failing.

Changes

• Added withEmptyMatchTreatment() method to AvroSortedBucketIO.Read
  and ParquetTypeSortedBucketIO.Read
• Updated SortedBucketSource to pass through the parameter to
  FileSystems.match() calls
• Maintains full backward compatibility (defaults to DISALLOW behavior)
• Added tests

Usage Example

import org.apache.beam.sdk.io.fs.EmptyMatchTreatment

val read = AvroSortedBucketIO
  .read[MyRecord](new TupleTag[MyRecord]("input"))
  .from("/path/to/potentially/empty/directory")
  .withEmptyMatchTreatment(EmptyMatchTreatment.ALLOW)

Testing

• All existing tests pass
• Added new test cases for the EmptyMatchTreatment functionality
• Verified backward compatibility with existing code

Closes #5759

[clairemcginty]

Hey, thanks for the PR! :)

I think this will need a few more changes to get it working end-to-end - it's more involved than it seems since the scio-smb implementation bypasses a lot of the usual Beam filesystem abstractions. Off the top of my head:

• BucketedInput#getOrSampleByteSize would need to be set up to handle empty input
• MultiSourceKeyGroupReader would need to be updated to skip empty directories when instantiating bucketedInputs

Adding tests in SortedBucketSourceTest/SortedBucketTransformTest should reveal these issues (and anything else I haven't thought of). Let me know if you need more pointers!

[sushmullur]

Thanks for catching that! Made the changes