Respect CDATA[[ sections when parsing HTML

[Corion]

This changes the HTML parser behaviour to properly respect
CDATA[[ sections and to ignore link tags in Javascript code.

The old behaviour can be restored by passing undef as the "marked_sections"
option when creating the WWW::Mechanize object:

my $mech = WWW::Mechanize->new(
    marked_sections => undef,
);

The patch also includes a (nasty) test file to check the old and
new behaviour

See also the discussion in https://perlmonks.org/?node_id=11116478 and https://gist.github.com/haukex/fd76efa16f0b07ce6a7441d9b2265b2a for more context

[coveralls]

Pull Request Test Coverage Report for Build 337

• 12 of 12 (100.0%) changed or added relevant lines in 1 file are covered.
• 41 unchanged lines in 1 file lost coverage.
• Overall coverage increased (+0.04%) to 94.279%

---

Files with Coverage Reduction	New Missed Lines	%
lib/WWW/Mechanize.pm	41	93.95%

Totals
Change from base Build 336:	0.04%
Covered Lines:	758
Relevant Lines:	804

---

💛 - Coveralls

[oalders]

@Corion would you be able to rebase this? I merged a much older PR just now that touches the same code.

[Corion]

No problem - the rebased version passes tests locally, let's see if it also passes tests on CI

[oalders]

Thanks @Corion!