Define plan for Thrust/CUB kernel unification

[jrhemstad]

Below is a list of tasks in prioritized order. We should start with algorithms that already exist in CUB. This will allow delivering CUB optimizations into Thrust sooner.

Replace Thrust Algorithms with CUB

Preview Give feedback

1. Refactor thrust::unique_by_key to use cub::DeviceSelect::UniqueByKey #1210
   +0
   
2. Refactor thrust::reduce_by_key to use cub::DeviceReduce::ReduceByKey
3. Refactor thrust::partition to use cub::DevicePartition
4. Refactor thrust::copy_if to use cub::DeviceSelect
5. Refactor thrust::unique to use cub::DeviceSelect::Unique
6. Refactor thrust/extrema.h to use cub::DeviceReduce

Port Thrust Algorithms into CUB

Preview Give feedback

1. Port thrust::cuda_cub::parallel_for to CUB
2. Port thrust::merge to CUB
3. Port thrust/thrust/system/cuda/detail/set_operations.h to CUB

A few notes:

• Partition and copy doesn't have stencil overload in CUB. We might experiment with using ::Flagged version along with a transform iterator.
• While porting thrust::reduce_by_key we need to decide on accumulator type.

[gevtushenko]

Updated issue description with the migration plan. Duplicated this plan in #26.