[GPU] Add basic GPU support and example

[matrix72c]

Basic GPU support for lowering linalg.matmul to GPU. Integrated from #290

• Create Buddy GPU Dialact, and add three ops: hoist_static_alloc , unroll_vectors_gpu_mma_sync, vector_to_mma_conversion for transform dialact inside GPU dialact.
• Extend MLIRGPU in Conversion, adding two passes: ConvertMemcpyToGPUPass and LegalizeShmemOutliningPass.
• Add tests for both ops and passes.
• Add a detailed example in examples/BuddyGPU.

[zhanghb97]

Please use clang-format to format all the .cpp/.h files.

[zhanghb97]

Thanks!