Filter out unsafe AI messages from the history

[SmittieC]

We now save unsafe AI messages. This needs to be filtered out from the chat history whenever we give it to the LLM to generate a response, as well as from the chat UI when debug mode is disabled.

Proposed approach:
We tag the offending messages with a new tag. Something like SAFTEY_LAYER_TRIGGERD. Whenever we send the history to th bot or show it to the participant, we filter out all AI messages with this tag, since it is unsafe.