Replace NDS2 data handling with a stage series of threads.

Data is passed in stages through threadsafe queues. This relieves a bottle neck in NDS2Client data handling where data was not requested until buffer pointers were incremented.

Also, conversion of data to single precision floats is now separated from channel callbacks and NDS2Client data reading.

Some changes made to improve abort times.