FEAT: Add supports_prefill capability and PrefillAttack strategy

[Dev3 ^~^ (shashank03-dev)]

Fixes #3270

Adds a supports_prefill target capability and a PrefillAttack that ends the conversation on an assistant turn carrying prefill_text, so the target continues it. The attack requires the capability natively, so a target that would drop the assistant turn raises at construction instead of looking like it resisted.

• HuggingFaceChatTarget sets the flag and now passes continue_final_message when the last message is from the assistant.
• OpenAICompletionTarget does not set it, even though the issue suggested completion-style targets. It forwards only the last message's text, so the user turn carrying the objective would be lost.
• The prefill is stored with the simulated assistant role, and request converters apply only to the objective.

Tests: unit tests for the attack, the capability round-trip and per-target defaults, and the HF chat template switch; the doc example is run against Qwen2.5-0.5B-Instruct.

[Manohar Paturi (ManoharPaturi)]

This is a great implementation of the prefill attack. Requiring supports_prefill natively in TargetRequirements is the right design decision because attempting to adapt or fall back on targets that drop trailing assistant messages would disguise a target limitation as an attack refusal. Scoping request converters strictly to the prepended user objective preserves the exact prompt tokens intended for continuation, and tagging the prefill as a simulated assistant role keeps memory history truthful. The HuggingFaceChatTarget chat template continuation tests cover the mechanics nicely.

[Dev3 ^~^ (shashank03-dev)]

Manohar Paturi (@ManoharPaturi) Thank you :)