webgpu: Dynamically load webgpu implementation at runtime
VTK could flexibly load available webgpu shared libraries at runtime instead of locking onto any specific implementation at compile time. This reduces the burden on engineering the build system to link against webgpu implementations without a proper CMake build system or none at all. Mozilla's wgpu-rs uses cargo
WebGPU C API offers wgpuGetProcAddress
for this purpose to query device function pointers at runtime. This is also supported by emscripten.