Add resumable download reports and Galaxy-safe recovery collections

[paulocilasjr]

Why this is important

mcdi download currently reports per-file results only in the job log. The Galaxy wrapper exposes only the downloaded-files collection.

If a Galaxy execution is interrupted:

• A new execution receives a new working directory.
• Previously downloaded files are not available to the new job.
• The same manifest is downloaded again from the beginning.
• Users do not receive a structured list of successful, failed, incomplete, or checksum-invalid files.
• Large GDC, PDC, or IDC cohorts can waste substantial time and bandwidth.

Expected behavior

The tool should produce:

• A visible collection containing successfully downloaded samples.
• A machine-readable download-state file, hidden in the Galaxy history by default.

Supported statuses should include:

• pending
• downloaded
• skipped
• error
• checksum_mismatch
• extract_error

Galaxy constraint

• Creating a hidden status report and retrying only unresolved files is feasible.
• Appending datasets directly to an existing Galaxy history collection by collection ID is not a normal or safe Galaxy tool operation.

The Galaxy-safe approach is:

1. Preserve the successful collection from the first run.
2. Download unresolved files into a new recovery collection.
3. Use Galaxy’s built-in Merge Collections operation to create a new combined collection containing the old and recovered datasets.
4. Leave the original collection unchanged.

The report should use a portable run_id and manifest fingerprint to associate related outputs. A Galaxy collection ID may be recorded as informational metadata if Galaxy can safely provide it, but retry logic must not depend on it.