Use slicing kernel to copy distances inside NN Descent

[jinsolp]

Description

This make use of raft's slicing kernel within NN Descent build.
I found that my previous implementation was inefficient (merged in this PR).

Improvements

Time to call NN Descent build() with return_distances=True before and after using this kernel

Dataset	Before	After
mnist (60000, 784)	1550ms	1020ms
sift (1M, 128)	11342ms	5546ms
gist (1M, 960)	13508ms	9278ms

[cjnolet]

LGTM!

[jinsolp]

If we try to use the mdspan copy, we get errors compiling.
graph_.h_dists has index type size_t, and the mdspan raft::copy complains if graph_d_dists index type is not the same.
But then slice complains about meaningless comparison with 0 if the index type is not signed, which makes me have to set it to int64_t when getting the const view of graph_d_dists.

I think it would be better to use the raft::copy I was previously using to keep things neat: )! @lowener @cjnolet

[lowener]

/merge