AnalogIO refactoring and support for buffering and data type conversions

[glopesdev]

• Adds support for buffering analog input and output data
• Adds support for optional data type conversion to/from Volts
• Introduce AnalogInputDataFrame and expose analog data using Mat types
• Fixes issue where input ranges were not being assigned to every channel
• Disallow reconfiguring IO channel direction at runtime
• Refactor all properties to remove ordering prefix

Fixes #79

[jonnew]

The order of samples in the buffer is currently incorrect. We need somehow to get each channel (C) sample from raw frames:

 [C0, C1, ..., C11] 

into the analog buffer with a BufferSize-sample (N) stride:

[C0_0, C0_1, ..., C0_N, C1_0, C1_1, ..., C1_N, ..., C11_0, C11_1, ..., C11_N]

Currently, data is ordered like this:

[C0_0, C1_0, ..., C11_0, C0_1, C1_1, ..., C11_1, ..., C0_N, C1_N, ..., C11_N]

The obvious ways of doing this will have a significant performance impact, I would imaging (e.g. looping over the data).

[glopesdev]

@jonnew oh, so it's actually a demuxed row major order? That's even better, I can just get rid of the transpose then, and it will be even more efficient!

[jonnew]

Aside from these issues, I'm encountering a strange behavior with my hardware. If I ground an analog input, I see a significant offset voltage (-1V or so). I'm not sure if this an issue with my hardware, hardware design, or HDL. If you can try to reproduce it would be helpful.

[jonnew]

I have verified that this now appears to be functioning well. There is an issue with the analog input, but I think its either hardware or gateware related and may even be specific to my prototype hardware. Its documented here: open-ephys/onix-fmc-host#8. Let me know if you see this -0.9V offset on grounded analog inputs on your hardware when you have time to test

[glopesdev]

Let me know if you see this -0.9V offset on grounded analog inputs on your hardware when you have time to test

@jonnew Just did the test this morning and I don't see this offset (there was a very slight 20mV offset but this could well be the harp device DIO we used for testing, as we were seeing this on the oscilloscope as well).