fix: resolve retrieval dataset corpus paths relative to training file

[oliverholworthy]

• Relative corpus paths in retrieval training JSON files are now resolve relative to the JSON file's directory, not the working directory
• This makes training data portable across machines and containers without needing to rewrite corpus paths

Changelog

• load_datasets() in retrieval_dataset.py: resolve relative path entries in the corpus config before passing them to add_corpus()
• Added unit test covering the relative path resolution

Tests

• New unit test test_load_datasets_resolves_relative_corpus_path
• All existing test_retrieval_dataset.py tests pass

Before your PR is "Ready for review"

Pre checks:

• Make sure you read and followed Contributor guidelines
• Did you write any new necessary tests?
• Did you add or update any necessary documentation?

If you haven't finished some of the above items you can still open "Draft" PR.

[oliverholworthy]

/ok to test f2938d7

[adil-a]

LGTM, thanks @oliverholworthy