Add workflow to run SDGym multi-table benchmark monthly and publish results

[R-Palazzo]

Problem Description

Once #515 is resolved, it should be possible to run the multi-table SDGym benchmark on GCP using enterprise synthesizers. The goal of this issue is to add GitHub Actions workflows that automatically run the multi-table benchmark and upload the results, mirroring the existing automation for single-table benchmarks.

Expected behavior

• Add a run_benchmark_multi_table.yaml file
• Add an upload_benchmark_results_multi_table.yaml

Additional context

The new workflows should closely follow the existing single-table benchmark workflows, with the following differences:

• Credentials handling:
  • Retrieve AWS, GCP, and SDV Enterprise credentials from SDGym GitHub secrets
  • Prepare these credentials in a format compatible with _benchmark_multi_table_compute_gcp
• Benchmark execution:
  • Use _benchmark_multi_table_compute_gcp to launch the benchmark runs
  • Launch:
    • One GCP instance running HMASynthesizer
    • One GCP instance running HSASynthesizer and IndependentSynthesizer
• Result handling:
  • Benchmark artifacts and results are uploaded to AWS (S3), consistent with the existing SDGym workflow