Increase operator coverage

[xuzhao9]

We would like to cover more operators from the following sources:

1. Tilelang benchmark: https://github.com/tile-ai/tilelang/tree/main/examples, deepseek_deepgemm, deepseek_mla, linear_attention

• gemm_sm100
• deepseek_deepgemm
• deepseek_mla
• linear_attention: https://github.com/meta-pytorch/tritonbench/pull/567/files

3. Mojo mha_sm100: https://github.com/modular/modular/blob/main/max/kernels/src/nn/mha_sm100.mojo

4. vLLM custom Triton kernels from MoE

[xuzhao9]

Need to add vLLM MoE kernel generated by TritonParse