Enable Vulkan

It would be cool to enable vulkan support (a bit like was done in https://github.com/conda-forge/llama.cpp-feedstock/pull/95) to permit the package to run on non-Nvidia GPUs, like Intel or AMD.

Long term it would be also cool to package ggml separately to share this packaging effort between llama.cpp and whisper.cpp (xref: https://github.com/conda-forge/staged-recipes/issues/29179), but to be honest it is not clear to me if ggml is stable enough for that, even if the Debian packaging of llama.cpp and whisper.cpp may be a good example to imitate, but for the time being just add vulkan support here would drastically enhance the usability of `whisper.cpp` conda-forge package.

fyi @pmorerio @conda-forge/whisper-cpp