Rather than implementing this ourselves (1+k lines of code, several large
data tables), this uses ZXing as an (optional) dependency, making the
implementation straightforward.

ZXing as a dependency is also interesting as it contains support for
all other formats as well, potentially allowing us to use it instead of
the other external dependencies. ZXing is larger than those, but it's
also used in other KDE code already, unlike qrencode and libdmtx.