Enable the large 2D load for dot operands

[chengjunlu]

The code is going to be submitted in several PRs:

• https://github.com/intel/intel-xpu-backend-for-triton/pull/1492
• https://github.com/intel/intel-xpu-backend-for-triton/pull/1512
• https://github.com/intel/intel-xpu-backend-for-triton/pull/1513
• https://github.com/intel/intel-xpu-backend-for-triton/pull/1537
• https://github.com/intel/intel-xpu-backend-for-triton/pull/1514
• https://github.com/intel/intel-xpu-backend-for-triton/pull/1516
• https://github.com/intel/intel-xpu-backend-for-triton/pull/1515
• Use the cluster size > 1 in accelerate matmul.
• unit test case changes for repCluster