Add vecpar implementations

[georgi-mania]

Provide implementations for scenarios

• [x] run in parallel using vecpar offloading lambda (CUDA and OpenMP)
• [x] run in parallel using single-source algorithms (CUDA and OpenMP)
• [x] ability to choose memory resource for GPU (managed-memory AND host-device memory)
• [ ] ability to choose different offloading backend (CUDA vs OpenMP target)
• [ ] scripts for automatic testing