Default GPU architecture

[Qubitol]

CUDA 13 drops the possibility to compile with compute capability 70 [1], that is the default we have set in CUDACPP.

I think we should make sure that the users understand that the way to fix it properly would be to specify the architecture explicitly with the environment variable MADGRAPH_CUDA_ARCHITECTURE.

Alternatively, we can think of making it dynamic by default.
One can get that number from:

nvidia-smi --query-gpu=compute_cap --format=csv,noheader

And then do something like this in the cudacpp.mk:

MADGRAPH_CUDA_ARCHITECTURE ?= $(shell nvidia-smi --query-gpu=compute_cap --format=csv,noheader 2>/dev/null | tr -d '.' | sort -nr | head -n1)

What do you think?

[1] https://en.wikipedia.org/wiki/CUDA#GPUs_supported