Currently we write out empty CSS classes (`class=""`) every time we
create a new tag. This adds 9 unnecessary bytes per span element. In a
recent trace, I have counted 11950 span elements. So we transported 105
unnecessary kilobytes!

This reverts merge request !29832

This change ensures that all open <span> tags are closed before adding a
<br> tag.

This makes ref validation to use always `UntrustedRegexp`.
This also splits the existing RubySyntax into separate
class.

RE2 differs from Ruby in handling multiple-line strings. The string "foo\n"
will not match the regular expression "foo$" unless multi-line mode is enabled
(and it's off by default).

Since we're already scanning the build trace line by line (and so multi-line
coverage regular expressions won't work), we can fix this by removing the line
separator before scanning the string.

24570 use re2 for user supplied regexp 9 3

See merge request !2129

Fixes https://sentry.gitlap.com/gitlab/gitlabcom/issues/27545/

of breaking ANSI sequence or Unicode

Given a valid pipeline job, and a regex which wouldn't match to a jobs
trace, the stream of the trace would return the File object. This was
not the case when it matched a value, as that would have been return
from the block.

Now the `extract_coverage` method returns `nil` if no match was found.