Rewrite port handler to support software-controlled disconnect & reconnect

[FilipDominec]

Double initialization of one device must be blocked.

The 2nd call somehow manages to connect to already connected device, but the identify callback
gets redirected to the first Rp2daq object:

In [2]: rp2daq.Rp2daq()                                                                                                                                       
2024-04-06 11:54:19,091 (MainThread) connected to rp2daq device with manufacturer ID = E66138935F6F8E28
Out[2]: <rp2daq.Rp2daq at 0x7f8dfee52c70>

In [3]: rp2daq.Rp2daq()                                                                                                                                       
2024-04-06 11:54:23,229 (Thread-273) Warning: Got callback that was not asked for
		Debug info: {'report_code': 0, '_data_count': 30, '_data_bitwidth': 8, 'data': b'rp2daq_231005_E66138935F6F8E28'}
2024-04-06 11:54:23,373 (MainThread)    port open, but device does not identify itself as rp2daq: b''
2024-04-06 11:54:23,374 (MainThread) Error: could not find any matching rp2daq device

So on line "[3]" we should have detected the device was not available, and reported an error instead.

Related to this: the port handler process must stop once we have sw-disconnect, then don't report hw disconnect and allow for clear reconnect.

In [4]: rp.quit()                                                                                                                                             
 In [5]: 2024-04-05 11:16:13,790 (MainThread) Device disconnected, check your cabling or possible short-circuits  

Also related: rp.quit() should release COM port on Win10, so that one can run it again.

And if there is a long command (like synchronous stepper_move) when the device gets physically disconnected, it hangs forever - because it ignores the disconnect event, and waits for the report over USB which never comes.

And when a device is still actively reporting some data (infinite ADC, gpio edge events etc.) and a script is re-started, it randomly won't initialize, because initialize immediately expects a specific short report. Maybe the device should be somehow reserved to stop any other reports (?), then port flushed, then called the init.