port to kded
the standalone binary was fairly awkward because it required its own private heap of a bunch of underlying libraries (e.g. xcb) that caused it to have a multi-MiB footprint. this excess consumption goes away by moving to kded since we now only have the private heap we actually require reducing the footprint to a couple hundred KiB
this sees a change in the dbus api since we cannot register the object on / any longer. instead we now register on an RDN-derived name.
moving forward we could opt to replace the dbus api with the builtin capabilities of kdedmodule (doesn't really have an advantage though other than getting rid of a fairly tiny class)