feat: add feedback command

[fmueller]

Motivation

• Introduce a non-destructive review step that evaluates a draft against a validated SEO brief and surfaces actionable editorial guidance without rewriting the draft.
• Provide both a human-readable Markdown report and a machine-readable JSON report to support manual editing and automation pipelines.
• Allow scoped reviews via section ranges and focused checks (seo, structure, clarity, style, evidence) to speed iterative editing.
• Keep outputs deterministic in structure and conservative about facts when an original --note is provided.

Description

• Add prompt templates in src/scribae/prompts/feedback.py to drive a structured, JSON-only feedback agent with an explicit schema.
• Implement the feedback engine and validated models in src/scribae/feedback.py, including context preparation, section extraction, prompt rendering, LLM invocation, and md/json renderers.
• Add the CLI entry scribae feedback in src/scribae/feedback_cli.py and wire the command into the CLI via src/scribae/main.py, supporting --body, --brief, --note, --section, --focus, --format, --out, --out-dir, --dry-run, --save-prompt, --project, --language, and --model flags.
• Add tests and fixtures: tests/unit/feedback_cli_test.py and tests/fixtures/body_multi_section.md to cover Markdown/JSON outputs, dry-run prompt printing, and section scoping.

Testing

• Ran uv run ruff check src tests and linting passed without issues.
• Ran uv run pytest and the test suite completed successfully with 120 passed.
• Started uv run mypy src tests in this environment but the invocation did not complete (interrupted); recommend running to completion in CI/locally.
• New CLI unit tests for scribae feedback (Markdown/JSON/dry-run/section behavior) were executed as part of the test run and passed.

---

Codex Task