• Fixed bug where nonexistent grammars would cause tokenizer to fail.
• Added mensbeam/html as a dependency, removed docopt/docopt and
ext-mbstring.
• Discovered bug when injections are removed from the stack when
tokenizing, investigating.
• Originally I had a concept of a readonly node tree for grammars with nodes owning other nodes thinking it would be necessary when tokenizing. It isn't, so they're more trouble than they're worth.
• "ownership" in Grammar\Reference objects is handled by an ownerGrammarScopeName property which is then used to get the grammar from the GrammarRegistry.