docs: add OWASP LLM Top 10 mapping tutorial

[SchmeatMilk]

Summary

Adds a new tutorial, docs/source/owasp_top10.rst, showing defenders how to run garak against a target and interpret results against the OWASP LLM Top 10 (LLM01–LLM10). Wired into the "Using garak" toctree in index.rst.

The tutorial maps garak probe families (e.g. prompt_injection, leakage) to the corresponding OWASP category, so findings can be reported to non-technical stakeholders in industry-standard language. It is framed for defenders / authorized red teams and includes a responsible-disclosure note.

Not duplicating an existing PR

Checked open PRs (gh pr list --repo nvidia/garak --state open); no open PR adds an OWASP Top 10 walkthrough. Related PRs (#1920 retag, #1856 intent stubs) are unrelated to this doc.

AI assistance disclosure

This contribution was prepared with AI assistance (SEB). The submitting human (Malik) has reviewed every changed line. Changes are documentation-only (a new .rst + one toctree entry); no code or test behaviour is altered.

Test commands run

• git status / git diff --stat confirm only the two doc files changed.
• .rst content reviewed for valid reStructuredText.
• No existing tests affected (doc-only change).

Co-authored-by: SEB (AI assistance)
Signed-off-by: Malik malik@seb.security

[jmartin-tech]

This is not the preferred path to activate OWASP probes. garak supports activating probes based on tags and owasp:llm?? tags exist on probes related to those weaknesses, also the --taxonomy option can be used to have the report html organized based on tags from an existing report.jsonl that can also be used for this use case.

Changes to how probes are selected are expected to land in main soon. Guidance on how to update this PR will be added once the feature branch that already includes #1831 merges to main.