Add before and after hooks for masking data in prompts

[vasinov]

Some users would like to mask and unmask data that goes out to external APIs as part the LLM prompt (e.g., customer PII). To solve this problem I propose adding the following to the framework:

1. Add BasePromptStackProcessor class that contains before_run and after_run methods. Let's think of better names.
2. Add prompt_stack_processors: list[PromptStackProcessork] property that contains a list of optional prompt stack processors.
3. Run PromptStackProcessor.before_run in BasePromptDriver.before_run and PromptStackProcessor.after_runinBasePromptDriver.after_run` in all prompt processors.
4. Add AmazonComprehendPiiProcessor that uses Amazon Comprehend to filter out PII data.
5. Add PromptDriverPiiProcessor that can use an arbitrary prompt driver to filter out PII data.

[ashkaaar]

can I work on this?