Add Anthropic Sonnet-powered LLM GIF generation pipeline

[apshoemaker]

Motivation

• Implement an LLM-driven producer that turns a natural-language prompt into an ND75-ready animated GIF so users can pipe generated animations into the existing push workflow.
• Use Anthropic (Sonnet) as the model backend and follow TDD: add failing tests first, then implement the feature to satisfy them.

Description

• Added nd75_screen/llm_gif.py to call the Anthropic Messages API, parse a JSON storyboard, render frames at 135x240 (ND75) and emit GIF bytes via generate_gif_bytes_from_prompt.
• Added a producer CLI nd75_screen.cli.llm (nd75-llm-gif) that accepts a prompt and writes generated GIF bytes to stdout for piping into nd75-push.
• Added tests/test_cli_llm.py with TDD-first coverage for CLI stdout behavior, default model wiring (DEFAULT_MODEL), rendering to ND75 size, and reading ANTHROPIC_API_KEY from the environment.
• Registered the anthropic>=0.84.0 dependency and the nd75-llm-gif script in pyproject.toml, updated README.md with usage examples, and updated the lockfile to include the new deps.

Testing

• Ran uv run pytest tests/test_cli_llm.py -q initially to confirm the red test (module missing) during TDD, then re-ran after implementation and observed 3 passed for that file.
• Ran the full test suite with uv run pytest tests/ -q and observed 79 passed indicating no regressions across the project.
• Verified the new CLI help with uv run python -m nd75_screen.cli.llm --help, which produced the expected usage output.

---

Codex Task