Use upstream warning functions for CUDA initialization

[charlesbluca]

Following up on dask/distributed#6121, this PR uses Distributed's warning functions for CUDA initialization, cutting down on code duplication.

cc @pentschev

[codecov-commenter]

Codecov Report

Merging #894 (2b2a48d) into branch-22.06 (952e130) will increase coverage by 24.47%.
The diff coverage is 0.00%.

❗ Current head 2b2a48d differs from pull request most recent head 062d3d4. Consider uploading reports for the commit 062d3d4 to get more accurate results

@@                Coverage Diff                @@
##           branch-22.06     #894       +/-   ##
=================================================
+ Coverage         65.01%   89.49%   +24.47%     
=================================================
  Files                22       16        -6     
  Lines              3076     2094      -982     
=================================================
- Hits               2000     1874      -126     
+ Misses             1076      220      -856     

Impacted Files	Coverage Δ
dask_cuda/initialize.py	87.23% <0.00%> (-0.27%)	⬇️
dask_cuda/benchmarks/local_cupy_map_overlap.py
dask_cuda/benchmarks/local_cudf_merge.py
dask_cuda/benchmarks/local_cudf_shuffle.py
dask_cuda/benchmarks/local_cupy.py
dask_cuda/benchmarks/utils.py
dask_cuda/_version.py

---

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 952e130...062d3d4. Read the comment docs.

[pentschev]

LGTM, thanks @charlesbluca .

[pentschev]

rerun tests

[pentschev]

@gpucibot merge

[jakirkham]

Also PR ( dask/distributed#6121 ) is part of the (currently in progress) 2022.4.1 distributed release