* Image::TIFF should not need to slurp
* convert the EXIF parsing code to just use something like
  Image::Exiftool - our code is way to complicated, outdated,
  and it's double work to maintain it
* Get rid of XML::Simple dependency in Image::Info::SVG::XMLLibXMLReader
* determining XBM files fail if a long comment precedes the "pseudo magic"
  Seen with /usr/local/include/X11/bitmaps/mouse.xbm. Possible solutions:
  * read more than 64 bytes for magic checks
  * make the file detection mechanism two-step: first step with a small
    buffer for file formats which have a real magic at beginning of the
    file, and if this fails following a second step for stuff like xbm