Increase support for prompt template building and transformation

[ohltyler]

A common user journey for a RAG use case is constructing a dynamic prompt template that takes input on-the-fly to pass to an LLM model (e.g., "summarize these found documents: ..."). Currently, users have to manually build the prompt templates from scratch, set as a sub-value in the model config, and blindly run to see the results. We should add a few enhancements:

• view the prompt template and injected values when constructing the input transforms Add optional prompt template transformation in input transform #379
• more presets and/or library of available prompt templates to select from, and a simpler way to configure