Sanitize input string or dict of strings by replacing sensitive data with placeholders.
It returns the sanitized input string or dict of strings and the secure
context as a dict following the format:
{
"sanitized_input":
The secure context is a bytes object that is needed to de-sanitize the response from the LLM.
