Encode.pm change to the UTF8 flag

Encode.pm change to the UTF8 flag

In the past, decode_utf8() has the effect of applying the UTF8 flag to
the decoded strings. However, Encode.pm changed around 2.53 and resulted
in wide-character errors preventing the likes of --load-all-sample from
working.

It seems that we can simply normalize the input to normalization form D,
then recompose it to form C on the way out, and sidestep the
decode_utf8() step entirely.

Signed-off-by: Dan Scott <dscott@laurentian.ca>