Provide `lower_bound` in `cuda.compute`

[shwina]

Specifically, I need vectorized lower bound: Specifically, I need vectorized lower bound: https://nvidia.github.io/cccl/thrust/api/group__binary__search_1gad2ac5a1c5091366144d6fdbf2ff5551e.html#lower-bound-exec-first-last-values-first-values-last-result

[griwes]

Binary search algorithms for c.parallel are now up for review: #7007