Title: Docs: Add content-fit guidance to the Slidev skill

1. Context & Motivation (Context)

• Issue: N/A
• Problem: The Slidev skill lists layout, Mermaid, transform, zoom, and code-scroll primitives independently, but it does not give agents a clear decision order for dense slides. That can lead agents to overuse whole-slide zoom or accept build-success as proof even when diagrams, tables, code, or text may overflow the visible canvas.
• Trade-offs: This keeps the change as additive skill guidance instead of changing runtime layout behavior. The trade-off is that agents get better heuristics without deterministic engine-level enforcement; that avoids API/runtime risk and keeps the PR scoped to the skill content.

2. Implementation Details (Implementation)

• Adds layout-content-fit.md as a dedicated reference for dense diagrams, tables, code, text, local transforms, slide zoom, and export verification.
• Updates SKILL.md so agents treat it as an index and load task-specific references for dense slides, Mermaid diagrams, export behavior, and advanced Markdown syntax.
• Cross-links layout-zoom, layout-transform, diagram-mermaid, and code-max-height to the content-fit workflow so the primitives are presented as complementary choices instead of isolated fixes.
• Updates the skill README capability list to mention content-fit guidance.
• Note: This PR intentionally does not change Slidev runtime, bundler, CLI behavior, or export output. Engine-level overflow detection would be a separate feature PR with tests.

3. Testing Strategies (Validation)

• Added documentation-only coverage for skills/slidev/references/layout-content-fit.md and linked it from the relevant skill references.
• Ran changed skill markdown link check for the modified skills/slidev files.
• Ran pnpm lint.
• Ran git diff --check.
• Verified backward compatibility with existing skill references: the change is additive and does not remove or rename existing reference files.
• Benchmark: N/A, docs-only skill guidance change.