Agent/conversation/request level temperature

[nalbion]

Currently the temperature is hard-coded to 1 - full creative mode.

This might be good for testing and some agents and prompts, but not every request. It should be configurable from a variety of sources:

• each Agent should have a default temperature
• each conversation should be able to over-ride the Agent's default.
• each function that calls create_gpt_chat_completion() should be able to over-ride the agent/conversation
• some models may need a maximum temperature (see also #120 )
• user should be able to define max/min/default temperature globally

[CRD716]

Also important for local models, ex. llama is generally run at 0.7-0.8, higher gets weird with code.