Compiles every token in a grammar (character-level: Literal, CharClass, Any, and RuleRef to another token -- the only node shapes an ordinary token body can contain) into one linked Grammar.VM.Program, run by Grammar.VM.Tokenizer.

A token whose entire body is a Grammar.IR.CustomLexeme (@native(...) at token position) is pulled out separately instead of compiled to bytecode -- there's nowhere for a {:call, name} from another token to jump into for it (see that module's own moduledoc for why it can't be composed inside a larger token expression or referenced from one), so Grammar.VM.Tokenizer dispatches to it directly by name via the second map this returns.