Add two new nodes that can be wired together to replace the original
NeuralBooru node, giving users control over LLM and validation steps
independently. The original NeuralBooru class is unchanged.
- llm_node.py: NeuralBooruLLM handles LLM communication (Ollama native
API + OpenAI-compatible), outputs (raw_tags, is_fallback)
- validator_node.py: NeuralBooruValidator handles tag validation +
template application, outputs (prompt, dropped_tags, tags)
- __init__.py: register new nodes via try/except ImportError block
- tests: 32 new tests (19 LLM + 13 validator), all 91 pass
- workflows/AI_Anime_split.json: example workflow using split nodes
