perf: update model and library for faster gemini call

prompt: fix gen ai tag escaping

prompt: modify gen ai prompt to output in html tag instead of markdown

test: revert back to gemini-2.5-flash

test: transformer and gemini-3.5-flash

test: use transformer and gemini-3.1-flash-lite

test: latency test with lstm model (transformers latency is 450-950ms excluding overhead)

feat: modify the generative ai prompt to focus more on the specified question

Update inference_api.py

init: upload required file for inference