Add additional Model Runs

[AnkitaNaik]

1. Adding the following models to the harness

• Qwen/Qwen3.5-397B-A17B-FP8
• Mistralai/Mistral-Large-3-675B-Instruct-2512-NVFP4
• meta-llama/llama-3-3-70b-instruct
• Qwen/Qwen2.5-72B-Instruct
• KimiK2

2. Add a script to automate the model runs - Script now runs the benchmark_runner as per the task. Retries failed domains thrice and runs the evaluator after the benchmark_runner finishes.

[anupamamurthi]

Missing usage block

also, is this a script for our own benchmarking? if so, can we move it into "scripts" maybe?

if the idea is to expose this to the users, we should add the right documentation etc.

[AnkitaNaik]

Good point. This is an internal usage script. Will move it under scripts.

[anupamamurthi]

looks good, requested small/minor changes