Fix range in Oniguruma regex for Hack grammar (#159)

[slevithan]

Fixes #159. See extended details there, but essentially, \xHH for values above 7F doesn't the work the same in Oniguruma (the regex engine used by TextMate grammars) as in other regex engines. As a result, a standalone \xff is an invalid UTF-8 encoded byte value, rather than a valid code point value as it would be if using \x{ff}. So although this regex isn't throwing in Oniguruma due to loose error handling for encoded bytes that are never used in a valid UTF-8 encoded byte sequence (bytes F5 through FF), this is in fact an invalid Oniguruma pattern.

This error is currently leading to edge case bugs in Oniguruma (code points above FF are not matched by this negated range) and preventing the Hack grammar from working with Shiki's JS engine (which has stricter error handling for this invalid Oniguruma pattern).

[salesforce-cla]

Thanks for the contribution! Before we can merge this, we need @slevithan to sign the Salesforce Inc. Contributor License Agreement.

[slevithan]

Heads up that there was a server error when I signed the CLA at the link above, and now when I return to the Salesforce contributor license agreement page and authenticate using GitHub, it tells me "You already signed the CLA on 2024-12-16".

[slevithan]

Created a new PR #163 to correctly trigger the cla:signed tag.