Created by: Sand3r-

This PR introduces several things:

• It extends existing Op Testing suite in C++ to cover Functional testing
• AVX2 JIT kernel for NCHW8C x NC is introduced and tests for it are provideed
• MKL-DNN Elementwise Mul has been rewritten and simplified (it's easier to analyse when checking the whole file than by diff changes now) and now support both AVX512 and AVX2 instruction sets
• x_format and y_format attrs have been removed from Elementwise Operator, resolves #16948 (closed)
• The Python UT for elementwise mul has been removed since it doesn't expose enough API to test this Op
• Instead, a C++ UT has been introduced that tests AVX512 and AVX2 implementations when the runtime architecture supports it.