Fix double-adding tokens in raw tags

[LiptonB]

The current logic for adding tokens in the raw tag has the potential to add to the body a substring of a token, and then the full token, resulting in doubled data. This happens when the token matches the FullTokenPossiblyInvalid regex, but does not contain the expected end tag. For example, this occurs when a template contains a string like "{% raw %}{{ {% test %} }}{% endraw %}", in which case the tokenizer produces "{{ {% test %}" as one of the tokens, and the template output is "{{ {{ {% test %} }}"