uhid fixes

A few minor changes, the 'big' ones are:

• UHIDDevice.destroy() polls all changes off the fd. previously it would only take the first which in my case was a mysterious open
• UHIDDevice.dispatch() loops until no further events are available. This means our udev events are actually somewhat synchronised now
• UHIDDevice.destroy() resets the internal device node/hidraw node list

There's a remaining issue in that the udev remove event will come after destroy() and not picked up until something else calls dispatch(). Largely a cosmetic issue while we clean out everything during destroy() anyway but it feels wrong. I played with adding extra dispatch() calls during destroy() but that makes things worse since we don't know what other events we may be handling.