[BUG] S3 Scan doesn't scan meta data when no filter is provided

Describe the bug
This config in the S3 Scan works fine

    - bucket:
        name: offlinebatch
        data_selection: metadata_only
        filter:
          include_prefix:
            - bedrockbatch/batch-input
          exclude_suffix:
            - .out

But without filter, the below doesn't work since no events (metadata only) are sent to the pipeline.

    - bucket:
        name: offlinebatch
        data_selection: metadata_only

To Reproduce
Steps to reproduce the behavior:
1.
Use the below config in the S3 Scan and an ml inference processor

  source:
    s3:
      acknowledgments: true
      delete_s3_objects_on_read: false
      scan:
        buckets:
          - bucket:
              name: offlinebatch
              data_selection: metadata_only

  processor:
    - ml_inference:
        host: "<your host url>"
        aws_sigv4: true
        action_type: "batch_predict"
        service_name: "bedrock"
        model_id: "<your model id>"
        output_path: "s3://offlinebatch/bedrockbatch/output/"
        retry_time_window: 
        aws:
          region: "us-east-1"
        ml_when: /bucket == "offlinebatch"

2. Once the pipeline started, the ml processor is not triggered and no logs showing the processor has received any events.
3. The data_selection: metadata_only might missed creating the metadata events in the S3 Scan if no "filter" is provided.