Vector-to-kernel pipeline bundle

[arun-thmn]

The PR Vector-to-kernel pipeline bundle does the following:

1. Bundles the four passes: brgemm tiling (Linalgtiling #974), vectorization (Linalg->Vector dialect pass and pipeline changes #958), hoisting vector transfers (Hoist vector transfers outside reduction and k loop #977), and vector to fma (Lower vector.contract to chain of vector.fma for matrix multiply. #978) together for the vector-to-kernel pipeline.

2. New json files are added to benchmark folder for benchmarking the vector-to-kernel pipeline.
   a) /base/vector.json: tests the base sequential version for mlp/gemm fp-32
   b) /omp/mlp-fp32-vector.json: tests the omp version for mlp/gemm-fp32
   c) /omp/torch-dynamo-vector-to-kernel.json: tests the omp vesion for torch mlp/gemm-fp32.

3. Updating the brgemm default tile size to lhsTile:4,32 and rhsTile:32,1, as this size emits the better gflops.

[adam-smnk]

Overall looks good

[arun-thmn]

@rengolin Responded to your comments. Please have a look.

[arun-thmn]

@rengolin Removed base/vector-to-kernel.json and moved the examples into base\base.json as you mentioned.

[rengolin]

LGTM, thanks!