Fix validation of translation unit section names

The diff is minimal. The issue was that \n was escaped to \\n by
preg_quote. When placed into a brackets in a regular expression,
it would instead match a backlash and the letter n. Because custom
unit names are rarely used, this issue would go unnoticed for a
while. When using a literal newline character, the issue goes away.

A better fix would be to encapsulate this in a method with tests,
but the fact that this is also used in a [^] context would require
more refactoring, which would delay fixing the issue.

Correct behavior can be verified manually using this code:
$ic = preg_quote( "_/\n<>", '~');
var_dump( preg_match( "~[$ic]~", "a\na" ) );
var_dump( preg_match( "~[$ic]~", '<foo>' ) );
var_dump( preg_match( "~[$ic]~", 'this-contains-letter-n' ) );
var_dump( preg_match( "~[$ic]~", "aa" ) );
var_dump( preg_match( "~[$ic]~", "a\\a" ) );

First two match (aka fail), while the rest do not.

Bug: T263546
Change-Id: I9bb3defcbac043160c813b1ebd7185b27be71f05