wp_read_image_metadata wrongly assumes IPTC metadata is Latin1

[koke]

Some applications like Aperture, use UTF8 in the IPTC tags. When exported images are uploaded to WordPress, the title and description metadata has the wrong encoding

The problem is in ​wp_read_image_metadata, since it runs utf8_encode for every field even if it's already utf8

To check if the parsed iptc is already UTF-8:

$iptc_is_utf8 = isset( $iptc['1#090'] ) && "\x1B%G" == $iptc['1#090'];

Depending on that result, the tags should go through utf8_encode or left as they are

Test script: dump IPTC contents

Test image: ​https://breakmeplease.files.wordpress.com/2012/04/dsc_4320.jpg

Title: Título

Description: Descripción

Related: #7495

Duplicate of #9417?

Seems that $iptc['1#090'] marker may not always be present. ​test-image-iptc.jpg (with IPTC tags written by IrfanView 4.33) doesn't have it. In that case, seems_utf8() check looks more reliable.