feat: add pixeltable skill for multimodal AI data infrastructure

[pierrebrunelle]

Summary

• Adds a pixeltable skill that teaches agents how to use Pixeltable for declarative multimodal data infrastructure
• Covers tables with multimodal column types, computed columns, embedding indexes, similarity search, UDFs, views with iterators, and 25+ AI provider integrations
• Includes reference docs for core API, agentic patterns, video/RAG workflows, anti-patterns, and provider catalog

Why this skill?

Agents can create tables, add computed columns that auto-execute on insert, build embedding indexes, and run similarity search without managing multiple tools. It handles images, video, audio, and documents natively.

Structure

skills/pixeltable/
  SKILL.md          # Main skill with frontmatter (name, description, triggers)
  README.md         # Summary and trigger keywords
  references/       # Detailed reference docs
    core-api.md
    agentic-patterns.md
    video-rag-agents.md
    workflows.md
    providers.md
    anti-patterns.md
    agents-memory-mcp.md
    ml-data-pipeline.md

Test plan

• test_skills_have_readme passes (README.md present and readable)
• test_catalogs passes (no regressions)
• Frontmatter follows existing skill conventions (name, description, triggers)