This matches the behavior of other browsers. We did the equivalent
change for <integer> in b86377b
We continue to store these as doubles for the extra precision.
Based on test-js-ast and test-js-bytecode, test-css-tokenizer finds CSS
files in CSSTokenizer/input, tokenizes them, and then dumps the output,
to compare against files in the expected/ directory. Our Tokenizer is
in a good state currently, but this will let us track regressions, and
also allow us to rewrite this in Rust and ensure that the output
matches.
To support test-css-tokenizer.py, we also have css-tokenizer.cpp which
exposes the Tokenizer, which is usually only used internally.
We'll want to expand this test suite over time, but this initial set of
tests covers each type of Token.
