Add pinned and as_cuda_array to cudasim mode

[tornikeo]

I'm working with a small cuda kernel project, and when testing my code with numba sim, I see that cuda.pinned has different arglist (single arg, instead of variadic *args) in these two modes for no good reason.

So, here's a small tweak to make it consistent.

[gmarkall]

gpuci run tests

[gmarkall]

Did this cause an error in your project? Or is this an aesthetic change only?

[tornikeo]

@gmarkall Not an error per se, but it forces me to monkeypatch the cuda.pinned. In a real run, I always pass several arrays in pinned. When I do testing, the pinned just errors out because of this. I'd expect a simulator to follow the real thing as closely as possible (in syntax and everything).

[tornikeo]

@gmarkall anything I should do for "Check Release Notes" failing?

[gmarkall]

@tornikeo If you'd like a mention of your fix in the release notes, you could add a release note snippet RST as documented in: https://numba.readthedocs.io/en/stable/developer/contributing.html#release-notes - however, as this is a small change, I'm happy if you'd prefer me to just mark this to skip the release notes check, and then I don't think there's anything further you need to do - let me know what you'd prefer.

[tornikeo]

@gmarkall Ah, it's a tiny change. No need to pollute the release notes with this. Thanks! 👍

[tornikeo]

@gmarkall somehow function as_cuda_array was missing from numba/cuda/simulator/cudadrv/devicearray.py. Ran into same issue when trying to test numba code on a gpuless machine. Added that function too.