TCPDriver MRO is broken about lantz

[hgrecco]

Thanks for taking the time to review this. You are completely right. I introduced this bug in 1d02773
Feel free to issue a patch against the develop branch and I will merge it.

The original idea was that all drivers were going to have a send and recv method. But send and recv (implemented in TextualMixin) called raw_recv and raw_send. So TextualMixin was before the "Raw" class to force calling TextualMixin.send and TextualMixin.recv

I now realize this is wrong as we are bitten many times. These are the things of the API that we realized by using it in different test cases. The plan is to have raw drivers that communicate with the instrument by sending and receiving bytes. They will ONLY raw_send and raw_recv.

On top, we build Textual based drivers that communicate with the instrument by sending and receiving strings and they implement send and recv.

So textual mixin goes at the end as you suggested. Thanks again.