Update Add Docker Support for CUTLASS FP8 GEMM

[SamirMoustafa]

This pull request includes changes to the cutlass_gemm kernel, aimed at improving the setup, installation, and usage of the kernel. The most important changes include the addition of a Dockerfile for containerized builds, updates to the readme.md for clearer installation instructions, and improvements to the setup.py for better path handling and CUDA library linking.

• kernels/cuda/cutlass_gemm/Dockerfile: Added a Dockerfile to facilitate building and running the project in a containerized environment. This includes instructions for building the image and running the container.

• kernels/cuda/cutlass_gemm/readme.md: Updated the README with detailed installation instructions for both Docker and non-Docker environments, including prerequisites and steps for building and running the project.

• kernels/cuda/cutlass_gemm/setup.py: Modified the setup script to dynamically determine the current location for include directories and to use CUDA_HOME for library directories, enhancing portability and ease of setup. [1] [2]

Minor Code Adjustments:

• kernels/cuda/cutlass_gemm/test_cutlass_gemm.py: Fixed import formatting for cutlass_scaled_mm to maintain consistency.

[SamirMoustafa]

I can also add a benchmark script that compares against PyTorch data types, but this would make the PR slightly longer. Please let me know if it is needed.