✨ 🍻 feat(question_eval): Integrate mimoracle dataset to question eval experiments

[honghanhh]

Fixes

Fixes #37

Description

This PR does

1. Extract the subset of mimoracle dataset with only meaningful sections:
   allergies|history of present illness|past medical history|discharge medications|social history|medications on admission
2. Integrate mimoracle dataset to question eval experiments
3. Log results - generate evaluation with 5 questions for the first 3 documents, each document with 6 above sections:

• gpt-4o

wandb: Run summary:
wandb:  mean/score/conformity 0.66667
wandb: mean/score/consistency 0.45556
wandb:    mean/score/coverage 0.11111
wandb:   sum/score/conformity 60.0
wandb:  sum/score/consistency 41
wandb:     sum/score/coverage 10

• gpt4o-mini

wandb: Run summary:
wandb:  mean/score/conformity 0.74444
wandb: mean/score/consistency 0.6
wandb:    mean/score/coverage 0.11111
wandb:   sum/score/conformity 67.0
wandb:  sum/score/consistency 54
wandb:     sum/score/coverage 10

Technical details

Definition of Done

• I followed the repository practices (e.g. I used gitmojis).
• I have added/updated tests according to my changes.
• I have added/updated documentation according to my changes.

[simonmeoni]

I have added some comments on separating the script into smaller methods and also considerations on adding the possibility to replace quickly the dataset we want to use using simple config in hydra

[RaphGardies]

LGTM for the parts I understand, I made a few minor comments