implementing GPU support to the codebase

[shivendrra]

To-Do's:

• Create a GpuArray class mirroring Array

• Integrate with GPU backends:

  • CUDA (for NVIDIA cards)
  • OpenCL / ROCm (for AMD)
  • (Optional later) Metal for Apple M-series

• GPU memory management abstraction

• Port CPU ops to GPU kernels:

  • Elementwise ops
  • Reductions (sum, mean, etc.)
  • Matrix multiplication & dot products

• Auto-select backend (CPU vs GPU) or allow manual selection

• Async GPU execution (streams, queues)

• GPU-CUDA kernel loader system

• Performance benchmarking against CuPy / PyTorch / NumPy

• GPU unit test framework

• GPU error handling and safe fallbacks

• Support for hybrid ops (GPU-to-CPU and vice versa)