Allow users to BYOLLM

[joshi4]

• export: Artifact -> Workflow
• config: support openai_base_url config value
• client: add llmClient that abstracts away the llm used to generate workflows
• client: refactor types to authz and llm pkgs
• client/llm: rm file
• llm,model,client: separate types for reuse
• llm: impl custom openAI endpoint
• enforce json output with structured outputs
• config,llm: validate llm_base_url and inherit model name from config
• refactor: move Ask api calls into llm pkg.
• custom_llm: impl Ask
• add requried fields
• tighten prompt
• support savvy explain with local llms