[FEA]: Initial hierarchy groups implementation

[davebayer]

Create an initial hierarchy groups implementation. We are interested in having:

• thread_group, a group that contains 1 thread,
• warp_group, a group that contains 1 warp,
• block_group, a group that contains 1 block,
• cluster_group, a group that contains 1 cluster and
• grid_group, a group that is an abstraction for the whole grid.

Each group should be specific to a given hierarchy.