This simplifies the logic for determining whether a token is reusable
and makes it more conservative. It should fix some incremental parsing
bugs that are being caught by the randomized tests on CI.
Previously, it was possible for references to external token states to
outlive the trees to which those states belonged.
Now, instead of storing references to external token states in the Stack
and in the Lexer, we store references to the external token trees
themselves, and we retain the trees to prevent use-after-free.
