[MCC-1547223] Python SDK - Add lazy frames to dataset versions

[dsilaghi-mdsol]

feature

• Add lazy frame access to get_dataset_versions(), matching R parity
• MCC-1547223

Checklist

• Review the pull request to fix typos and ensure variable/function names are intuitive, etc.
• Make sure the pull request does not have excessive number of unnecessary commits.
• Make sure you have added unit tests for the code changes. Tests should be added in the tests/ folder.
• Modify docs, if required. Add any new documentation in the doc/ folder.
• Rebase on latest active development branch (develop/main).

Changes Summary

Add a frame field to DatasetVersion and attach a lazy DatasetFrame in get_dataset_versions(), bound to each returned version's own dataset_uuid. This lets callers fetch an older version's data via version.frame.head() / version.frame.collect(), matching the existing Dataset.frame behavior and R's per-version frame attachment. No data is fetched during listing; existing metadata, sort order, and error handling are unchanged.

Adds regression tests covering: lazy fetch on demand, correct UUID routing per version, row limits (head/collect), field-set/equality/hash/repr compatibility with frame=None construction, empty results, and fetch/closed-client error propagation.

Deployment Notes

No API breaking changes — frame is an additional field with compare=False, so existing equality/hash/construction of DatasetVersion is unaffected. Requires a version bump on main for consumers to receive this via a published package (see prior investigation: v1.1.0 tag already existed before this branch's parity work merged, so the release pipeline skipped publishing a new build).

[ibaig-mdsol]

Closing per David and Robert.