Try out improving study and experiment handling

[bedroesb]

• Added a richer ENA integration test fixture and switched CI to use it, so the mars-cli workflow now exercises two source/sample chains, two ENA experiments, one paired-end run, and one single-end run instead of the earlier single-file happy path.
• Updated the ISA generation helper to rewrite all assay dataFiles in the selected template, not just the first assay entry, and to generate matching unique FASTQ files and checksums for each of them.
• Kept the PoC submission prep flow aligned with that new generator behavior, so the generated poc_work/isa.json and uploaded test data now reflect the full multi-file assay structure.
• Adjusted the CI workflow to point at the new multi-file ISA template, making the repository-services integration test cover the expanded ENA submission scenario end to end.
• Fixed ENA accession writeback behavior in the updated ISA JSON so assay-level ENA accessions are written only on the assay, while study-level ENA accessions stay on the study instead of being duplicated across both levels.
• Updated the related Python tests to match the new accession writeback behavior.