Revert "model mapping fixes for broken api endpoints etc"

model mapping fixes for broken api endpoints etc

requirements file update

model additions

fixes to load env vars properly, bugged out before and got annoying

Add python-dotenv dependency to requirements.txt

Add latest AI models: Claude 4 (Opus/Sonnet + thinking), OpenAI (o3/o3-pro/o4-mini), Gemini 2.0/2.5, Grok 3 thinking variants, DeepSeek R1-0528, new Mistral models

add atomic file writes to prevent results.json corruption

fix import errors when API keys are missing, add clearer error messages

add requirements.txt with pinned dependencies

fix max_tokens not being passed to API calls

timeout and token limit research; empirical analysis of processing times and token usage patterns, timeout testing experiments, discovery that historical slow scenarios now complete quickly

time limit experiment analysis and results; statistical experiment comparing model performance over time, a pilot thinking experiment with timing baselines, results dashboard and analysis docs, timeaware model configs and prompts

grok3 mini reasoning levels

add 4.1 to results