Optimized padim

[YToleubay]

added enable_optimization and compare_optimization flags to use optimized code and to compare the results of the optimized code with the original code.

added embedding_concat_optimized that uses PyTorch and performs checkerboard pattern downscaling and concatenation.

added embedding_concat_numpy that uses NumPy and performs checkerboard pattern downscaling and concatenation.

added postprocess_optimized

added infer_optimized that works through vectorization

added infer_init_run to run padim with dummy data to ensure hot inference later

On RTX 4060 Laptop GPU, Optimized code infers default ./bottle_000.png in average 10.25 ms while original code takes 203.0 ms