Core structure updates: image and text featurizer

[zzachw]

Core structure updates to support image and text data. I will elaborate more on the changes in tonight's meeting.

Framework:

• Re-design base and sample dataset to handle multi-modality data
• Change the task from a function to a class to store more information
• Add featurizers for image and value

Dataset and task:

• Add Covid19 CXR dataset and task (image classification)
• Add Medical Transcriptions dataset and task (text classification)

Model:

• Add ResNet model for image
• Add supports for models from hugging face's transformers library.

Note: PR is not finished yet. I will need to add more comments and documentation.