Require parameters for llm drift evals

When implementing `LLMetric` as part of `LLMDriftProfile` logic, we'll need to make sure that we add logic to check for bound parameters in evaluation prompts. It should be expected that a combination of `input` or `output` parameters are provided in the prompt.

This is needed because when the server is running LLM evals with the user-defined workflow, it will inject the input and output into the prompts before calling and returning a metric.