FEAT: Garak sysprompt extraction scenario

[rlundeen2]

Ports garak's sysprompt_extraction.SystemPromptExtraction probe as a PyRIT scenario, operationally similar to the garak probe but built with scenario constructs.

What it does

• Installs a real system prompt (sampled from the garak_drh_system_prompts / garak_tm_system_prompts datasets) on the target, then sends an extraction request asking the model to reveal it.
• Extraction templates live in a new local garak_system_prompt_extraction dataset; each template carries a per-seed technique tag mapping to one of the 9 garak categories, which become SystemPromptExtractionStrategy members.
• Each selected category becomes one AtomicAttack whose seed groups are (system prompt x template) combinations. Across all selected categories the total combinations are randomly sampled down to prompt_cap (garak's soft_probe_prompt_cap) so a default run stays bounded.
• Responses are scored deterministically by SystemPromptExtractionScorer (character n-gram containment overlap, reusing ApproximateTextMatching), wrapped by FloatScaleThresholdScorer.

Notes

• Follows the web_injection scenario pattern for dataset loading/resolution.
• Docs (doc/scanner/garak.py / .ipynb) updated with a minimal executable run.
• Added derczynski2024garak to the bibliography.
• Tests: 23 passing; ruff + ty clean.