Add MedLingo dataset + jargon expansion task (with tests, docs, example)

[anuragraychowdhury]

Summary

• Add MedLingoDataset implementation + config YAML
• Add MedLingoJargonExpansionTask for prompt/label construction (supports both shot modes)
• Add fast synthetic unit tests for dataset + task (tmp dirs, no network)
• Add API docs stubs and index toctree entries
• Add an example script demonstrating a default synthetic smoke run and an optional TransformersModel forward pass behind an env flag

Test plan
PYTHONPATH="$(pwd)" pytest -q tests/test_medlingo_dataset.py tests/test_medlingo_jargon_expansion_task.py
PYTHONPATH="$(pwd)" python examples/medlingo_medlingo_jargon_expansion_transformersmodel.py

Notes / scope vs paper
This implementation treats the task as multiclass classification over answer strings; the paper’s MedLingo evaluation is open-ended generation (often judged with an LLM), so metrics are not expected to match that protocol out of the box.

Original Paper: https://arxiv.org/abs/2505.15024
examples/medlingo_medlingo_jargon_expansion_transformersmodel.py