[REDO] unified handling for VectorMask object re-materialization during de-optimization

Currently mask values are propagated using vector registers, shape of mask vector matches the shape of the operands of a vector operation.
Objects allocated in the compiled frame are re-constructed during de-optimization.

To provide a unified handling for VectorMask instance reconstruction when mask is held either in a predicate register (for Intel AVX512 or ARM SVE) or vector register.