Boilerplate metrics

As I understand, the main goal of this is to easily express algorithms that can use either standard dispatch or generated functions, depending on what's known statically. Writing very general high-performance code for a given situation is always possible; the problem is that it can sometimes require lots of boilerplate.

Would it be helpful to make this explicit? A given proposal could be weighed based on its effect on reducing boilerplate for high-performance code that works for both static and dynamic cases. Maybe there could even be a small collection of running examples. Complete code isn't necessary for this, what matters is dispatch patterns and efficiently (in terms of human time) passing the right information for generated functions.

No files, tests, or entry points are named. Start by reviewing the stated goal of reducing boilerplate for algorithms using standard dispatch or generated functions, then identify representative dispatch patterns and running examples. Done means the proposal makes these patterns explicit and evaluates how efficiently the right information reaches generated functions.