Don't throw fatal exception during 'import cudaq' if missing dependencies

[bmhowe23]

If the user has GPUs but doesn't have all the right dependencies installed, we should not throw fatal exceptions in that case. Rather, we should fallback on qpp-cpu and log an info message that we are not selecting the nvidia target because dependencies are missing.

Helps mitigate the impact of missing dependencies as encountered in #1096.

Testing this in CI is difficult so I did standalone testing as follows:

First perform the following to force missing dependencies:

$ mv /usr/local/cuda/targets/x86_64-linux/lib/libcublas.so.11 /usr/local/cuda/targets/x86_64-linux/lib/libcublas.so.11.ignore

Without this change, it behaves as follows:

$ python3 -c "import cudaq"
Traceback (most recent call last):
  File "<string>", line 1, in <module>
  File "/usr/local/cudaq/cudaq/__init__.py", line 28, in <module>
    from ._pycudaq import *
ImportError: Invalid simulator requested: custatevec_fp32

With this change, it behaves as follows:

$ python3 -c "import cudaq"
<no errors>

[khalatepradnya]

👍

[github-actions]

CUDA Quantum Docs Bot: A preview of the documentation can be found here.