[REVIEW] OPG pagerank (CUDA)

[afender]

1D OPG pagerank checkpoint for the CUDA part

• accepts a regular CSC graph locally with partitioning offsets from [REVIEW] OPG pagerank (Py) #944
• gets comm info from RAFT handle
• uses cusparse spmv

There are still several things to do for OPG PageRank :

• More tests and analysis
• Get the result back to dask cudf
• Library-wise handle strategy [FEA] Accept RAFT handle everywhere #815
• Upgrade to raft cusparse wrapper once upgrades for cuda 11 are completed

Close #816, progress on #485

[GPUtester]

Please update the changelog in order to start CI tests.

View the gpuCI docs here.

[codecov-commenter]

Codecov Report

Merging #947 into branch-0.15 will not change coverage.
The diff coverage is n/a.

@@             Coverage Diff              @@
##           branch-0.15     #947   +/-   ##
============================================
  Coverage        48.89%   48.89%           
============================================
  Files               58       58           
  Lines             1579     1579           
============================================
  Hits               772      772           
  Misses             807      807           

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 9f71b90...f9671c7. Read the comment docs.

[Iroy30]

LGTM. Added some wrapper edits, opg pagerank e2e runs succesfully for basic test.

[afender]

rerun tests