Create a libcuda-stub library

[isuruf]

Because we've enabled error_overlinking on staged-recipes and since /usr/local/lib/libcuda.so is outside of prefix, conda-build errors. Correct way to fix this would be to generate the stub library libcuda.so ourselves and make that a CDT.

[kkraus14]

Note there's licensing terms about distribution driver libraries: https://docs.nvidia.com/cuda/eula/#attachment-a

The NVIDIA CUDA Driver Libraries are only distributable in applications that meet this criteria:

1. The application was developed starting from a NVIDIA CUDA container obtained from Docker Hub or the NVIDIA GPU Cloud, and
2. The resulting application is packaged as a Docker container and distributed to users on Docker Hub or the NVIDIA GPU Cloud only.

In addition to the rights above, for parties that are developing software intended solely for use on Jetson development kits or Jetson modules, and running Linux for Tegra software, the following shall apply:

• The SDK may be distributed in its entirety, as provided by NVIDIA, and without separation of its components, for you and/or your licensees to create software development kits for use only on the Jetson platform and running Linux for Tegra software.

[isuruf]

I was suggesting creating the stub library ourselves so that we don't have to use NVIDIA's. It's just a stub library, so it just needs the symbols, not the implementations.

[kkraus14]

@jakirkham would it be possible for you to poke the nvbug or someone internally to see if conda-forge could be granted permission to ship the libcuda.so stub library? The stub libraries in general seem like a gap in the EULA and the hoops we're jumping through as far as symlinking to avoid shipping it are getting a bit ridiculous.