Fix UTF-8 handling in VimHTML

[mateu]

This fixes UTF-8 mojibake in Pod::Elemental::Transformer::VimHTML when highlighted input arrives as already-decoded Perl text.

In the failing case, content like guillemets in #!vim perl blocks can render as «...» instead of «...».

The fix keeps the change narrow:

• normalize likely double-encoded-looking decoded input before passing it to Text::VimColor
• avoid pre-encoding already-decoded Perl strings to octets
• only decode HTML output if it is not already flagged as UTF-8
• add an explicit use Carp qw(confess); import while touching the module

Downstream context:

• Fix UTF-8 mojibake in #!vim highlighted blocks perladvent/Perl-Advent#663
• Add VimHTML transformer with UTF-8 handling fix perladvent/Pod-Elemental-Transformer-SynHi#3
• VimHTML can double-encode UTF-8 when input is already decoded Perl text #2

Happy to adjust the shape if you’d prefer a different approach.

[mateu]

I added a narrow upstream regression test for this too:

• t/vimhtml_utf8.t

It exercises Pod::Elemental::Transformer::VimHTML directly, checks decoded Perl text and UTF-8 octet input, and asserts we preserve guillemets without leaving «/» mojibake markers behind.