Add more open weights models to the Evals

[tomschelsen]

https://roocode.com/evals says requests should be made on Discord but can't join, so here would be my wishlist. Those are mainly smaller models than those currently in the evals, and so would cover a new part of the spectrum, for people who can neither use external services nor dedicate a full 8xBig_Gpus node :

• openai/gpt-oss-120b
• mistralai/Devstral-Small-2-24B-Instruct-2512
• nvidia/NVIDIA-Nemotron-3-Nano-30B-A3B-BF16
• Qwen/Qwen3-Coder-30B-A3B-Instruct
• openai/gpt-oss-20b
• mistralai/Devstral-2-123B-Instruct-2512

Thanks