feat: Add support for inference parameters

[hiro-v]

Problem

• I need to use Nitro as API for LLM
• Add support for using temperature, max_token, penalty in API call
• Add stream False
• Add stop

Success Criteria

• Fully supported

Additional context

• Check from LMStudio

[dan-menlo]

@vuonghoainam - Btw, the screenshot is broken

[tikikun]

These below params will be added

frequency_penalty
presence_penalty
logit_bias
stop