This logic decides whether to:

* dup an incoming keyword arguments hash
* replace the original argument with the dup
* set or clear ruby2_keywords_hash flag
* use the original, dup, or nothing for kwargs processing

The conditions leading to these various actions were intermingled
with performing the actions, making it difficult to adjust that
logic without breaking unrelated code. The new logic defines an
enum of all combinations of behavior and provides a straightforward
set of if/else branches to choose an action. This allows both the
interpreter and the JIT to use the same conditional logic but with
each handling the actions in an appropriate way (e.g. JIT may not
have an arguments array but it can replace fixed-arity positional
arguments).

First stage of refactoring keyword arguments logic to unify
interpreter and JIT and make it easier to follow and maintain.

This follows the refactor of variable-arity varargs logic, shared
by both JIT and interpreter. With specific arity, we never have
an action that replaces the incoming argument, nor do we return
UNDEFINED or clear the ruby2_keywords_hash flag.