feat(probes): Add multi-turn GOAT probe

[saichandrapandraju]

This PR introduces the GOAT (Generative Offensive Agent Tester) probe, a multi-turn red-teaming system that uses an Observation-Thought-Strategy-Reply (O-T-S-R) reasoning framework to iteratively craft adversarial prompts.

Verification

List the steps needed to make sure this thing works

• Supporting configuration such as generator configuration file

{"openai": {"OpenAICompatible": {"uri": "https:<placeholder>/v1", "model": "qwen2", "api_key": "DUMMY", "suppressed_params": ["n"], "max_tokens": 512, "temperature": 1.0, "top_p":null, "frequency_penalty": null, "presence_penalty":null, "seed":null, "stop":null}}}

• garak --model_type openai.OpenAICompatible --model_name qwen2 --generator_options '{"openai": {"OpenAICompatible": {"uri": "<placeholder>/v1", "model": "qwen2", "api_key": "dummy", "suppressed_params": ["n"], "max_tokens": 512,"temperature":1.0, "top_p":null, "frequency_penalty": null, "presence_penalty":null, "seed":null, "stop":null}}}' --probes goat.GOATAttack --probe_options '{"goat": {"GOATAttack": {"max_iterations": 5, "attacker_model_type": "openai.OpenAICompatible", "attacker_model_name": "qwen3", "attacker_model_config": {"temperature": 1.0, "uri": "https:<placeholder>/v1", "api_key": "dummy", "model": "qwen3", "max_tokens": null, "top_p": null, "frequency_penalty": null, "presence_penalty": null, "seed": null, "stop": null, "stream": true}}}}'
• Run the tests and ensure they pass python -m pytest tests/
• Verify the thing does what it should
• Verify the thing does not do what it should not
• Document the thing and how it works (Example)

[github-actions]

DCO Assistant Lite bot All contributors have signed the DCO ✍️ ✅

[saichandrapandraju]

I have read the DCO Document and I hereby sign the DCO

[saichandrapandraju]

recheck