CDA/FHIR connector components

[jenniferjiangkells]

Description

Refactors CDA/FHIR parsing components as part of the pipeline instead of in the sandbox use case.

Related Issue

#63 #71

Changes Made

• Added Connectors modules to pipelines and removed parsing code from cds and clindoc sandbox functions. This gives more visibility and explicitness to the step that converts request/response data to a pipeline container like Document
• Added methods .add_input() and .add_output() to BasePipeline and modified __call__ method to handle additional connector logic
• 💥 Breaking change: renamed .add() method to .add_node() to make it more explicit and follow the convention of the connector add methods
• 💥 Breaking change: renamed modules that contain Base classes to base.py within the pipeline and component modules for uniformity and brevity when importing. i.e. from healthchain.pipeline.base import BasePipeline
• Add __init__ to fhir_resources module as need to import Bundle elsewhere
• Added validate_and_convert_resources() model validator to Bundle, which allows Bundle to be initialised directly on a dictionary - allows dynamic import of resource type and recursively converts nested dictionaries to the appropriate Pydantic models e.g. Patient
• Added helper methods to CdsFhirData such as .create(), which initialises the model on a copy of the dictionary to avoiding modifying the original (this is due to the Bundle validation method potentially modifying the input dict)
• Added empty LLM and SummarizationPipeline and connectors to MedicalCodingPipeline, which are works in progress
• Updated documentation

Testing

• Added tests for all new components and functionalities, with the exception of the Bundle validator - @adamkells wanted to get your opinion that this is ok first.

[adamkells]

LGTM

[adamkells]

Will this class still be relevant with the Spacy, HuggingFace etc. transformers?

[jenniferjiangkells]

should they inherit from Component?