Add Tiny Tile support to CCL OPs

[jvegaTT]

The objective of this issue is to create the infrastructure needed and modify the implementations so that tiny tiles (i.e. tiles with a size smaller than 32x32) can be supported by the AllGather, Reduce-Scatter, and Barrier OPs. The objectives are below:

• [x] Modify the testing infrastructure to support tiny tiles and add tiny tile tests to all three OPs
• [x] Perform the necessary changes to the tensor manipulation libraries required by CCL to support tiny tiles
• [x] Modify the host and kernel code to take tiling information from the Tensor itself
• [x] Fix the all gather code and libraries it calls to make tiny tiles work
• [ ] Fix the reduce-scatter code and libraries it calls to make tiny tiles work
• [ ] Validate that the results are correct regardless of the tiling pattern used

[jvegaTT]

Re-scoped to only modify tile size in the h direction to comply with hardware limitations.