Add big batch sample dataset ingestion and benchmark tool

[tuannx]

What this PR does / why we need it:

Adds create_sample_big_batch_dataset.py for generating and ingesting large batches of sample files (e.g. hundreds or thousands of files) for Dataverse performance testing and benchmarking.

Motivation & Context:

Currently, create_sample_custom_dataset.py allows generating N files into the "Dataverse performance test dataset", but relies on CairoSVG (which requires system C libraries like cairo-2) and generates large PNG files on disk. For evaluating large-scale dataset ingestion (such as testing batch uploads of 500, 1,000, or 10,000+ files as discussed in community reports and IQSS/dataverse#12728), generating thousands of heavy image files is impractical.

create_sample_big_batch_dataset.py provides:

1. Lightweight file generation: Quickly creates dummy text or minimal 1x1 image payloads without heavy external C-library dependencies.
2. Batch upload pacing: Ingests files in configurable batch sizes with inter-batch delay (replicating tools like DVUploader or bulk migration pipelines).
3. Latency & throughput metrics: Reports real-time speed, min/max/average latency, and compares the first 10% vs last 10% of files to evaluate ingestion scaling.
4. Local file mode: Supports --generate-only to write files to disk for consumption by create_sample_data.py.

Related issues/PRs:

• Relates to performance testing in IQSS/dataverse#12728 and IQSS/dataverse#12729.

[pdurbin]

@tuannx thanks for the PR! I mentioned this morning during tech hours in the context of this related PR you made:

• 12728: Fix O(N) database updates during file upload on large datasets dataverse#12729