Refactor VHDL N-Dimensional Array Access

[leftink]

@ktbarrett, I wanted to propose a change for next Major release that would break backwards compatibility but get rid of some weird code I wrote early on.

For multi-dimensional arrays in VHDL, and I am not referring to arrays of arrays but true multi-dimensional arrays, you need to provide all dimensions to get an element in the array. When I implemented this capability, due to some limitations in the existing GPI interface and also my limited experience with python at the time, I treated the arrays as arrays of arrays. For example, to access an element in a 2-dimensional array the python interface would be:

element = arr[0][1]

While this worked, there was some "hacky" code written in C to manage this, not sure if it is still this way, but what I remember is having to maintain Pseudo handles that stored the first index because all indices were necessary to actually get a handle to the element of a multi-dimensional array. If I remember correctly, FLI could handle the current way of accessing multi-dimensional arrays, but VHPI required all indices to be provided.

Anyway, I propose we make multi-dimensional array access in python look the same as it would in VHDL and mimic the format used by numpy's ndarrays, i.e.

element = arr[0,1]

If you don't think it is worth it that is fine but wanted to throw it out there for consideration. I was never a huge fan of my approach to use Pseudo handles (which is also used on Generate loops) since it is a "fake" handle. Maybe we just need to take a look at all the Pseudo handle stuff as I am sure there is a better way than what I did there so long ago.