Bake the MiniLM model into the Docker container for deployment

[karlhigley]

This reworks the context generation code to:

• Load the MiniLM model from a local copy baked into the Docker container (instead of downloading from HF)
• Create the OpenAI client inside the ContextGenerator component

And also addresses some issues with deployment:

• Remove the ContextGenerator component from other pipelines to avoid loading the MiniLM model multiple times
• Bump the OpenAI version to account for a breaking change in the underlying dependency httpx
• Increase the amount of memory and ephemeral storage allocated to the Lambda function

[zentavious]

Solved the merge! Thank you for making these changes.