v0.15.0

• Refreshed the model defaults (EVOLUTION-PLAN item 4). Chat goes from llama3.2 /
  llama3.2:1b to qwen3:4b and qwen3:1.7b (the smaller named model): qwen3 is what
  the official quarkus-langchain4j Ollama guide uses, and its tool-calling is markedly better —
  which is precisely what the tool and agent templates exercise. The in-process embedding model
  goes from langchain4j-embeddings-all-minilm-l6-v2 to
  langchain4j-embeddings-bge-small-en-v15-q, upstream Easy RAG's recommendation, with a
  better MTEB score at a similar footprint. Both artifacts stay managed by
  quarkus-langchain4j-bom (no version pin); a local build against platform 3.38.1 resolved the
  new one at 1.17.2-beta27 and compiled green. Touches application.properties.template,
  pom.xml.template, RagSetup.java.template, the scaffold-project RAG section, and
  docs/VALIDATING-TEMPLATES.md.
• Checked and kept model-id. The Ollama guide appears to show chat-model.model-name, which
  looked like a wrong key in our templates. It is not: "model-name" in the reference is the
  placeholder for a named configuration, and the config model generated by
  quarkus-langchain4j-ollama 1.12.2 lists chat-model.model-id as the real property, with no
  chat-model.model-name alongside it. No change — recorded so the question is not re-opened.