Implement Webgpu device

[favilo]

Tried an initial implementation for #604

Based on the tasks from #597

1. Add webgpu feature flag

2. Create an Webgpu device struct under src/tensor/webgpu/device.rs & add impl DeviceStorage for WebGpu

3. Implement tensor allocation methods under src/tensor/webgpu/allocate.rs. See src/tensor/cuda/allocate.rs for examples

   1. impl ZerosTensor
   2. impl OnesTensor
   3. impl OneFillStorage
   4. impl SampleTensor
   5. impl CopySlice
   6. impl TensorFromVec
   7. impl TensorToArray

4. Implement skeleton kernels for Webgpu under tensor_ops - tests should compile & run but fail

   1. Add module webgpu.rs to all kernel folders
   2. Add a skeleton impl for all kernels (put todo!() in every kernel method)
   3. add impl Device<E> for Webgpu to src/tensor_ops/utilities/device.rs

5. Actually implement OpenCL kernels - this is highly parallelizable, each tensor op can be worked on separately. Every tensor op with a folder will need a kernel

[coreylowman]

@favilo this is AWESOME!! I'm thinking we can merge this into main without actually implementing any of the kernels here, and then people can start adding them if they want. Does that sound good, or did you have plans to add more to this PR?

[favilo]

I'm good with that, or you can make a new branch and we can add to that one. I figured we'd want to start adding kernels in other PRs mostly because I'm not entirely sure where to start with them.

So I'm pretty much done with this particular code.

[favilo]

Rebased onto main, again.

[coreylowman]

@favilo 🎉🎉🎉