feat: `create_evaluator` decorators (#3642)

* Spike out generic evaluator decorator

* Flesh out type annotations

* Add informative signature bind error messages

* Add tests for decorator evaluators

* Improve binding magic

* Update tests

* Ruff 🐶

* Improve error message

* Un-nest `results` from output if possible

* Ruff 🐶

* Use more robust output un-nesting

* Use updated kwarg

* Swap evaluator argument order

* Finish swapping evaluator order

* 'experiment_input' -> 'input'

* Change validator to allow `input` parameter name