Replace the regex/escape-normalizer with @csstools/css-tokenizer (already
present transitively). Tokenizing resolves comments, hex escapes, and line
continuations into the real value before scanning, so obfuscated forms (a
scheme split by a line continuation or hex escape) cannot slip past. Scan
url-tokens and string-tokens to cover url(), quoted @import, and string-form
image functions. Add regression tests for the bypasses.
Reject status page custom CSS containing @import or a url() / image-set()
/ image() / cross-fade() whose target is http(s) or protocol-relative,
while allowing data: and relative URLs. The check normalizes comments and
CSS escapes first so obfuscated forms cannot slip past. Enforced in the
create/update body validation.
Buco7854