[REVIEW] Updates to support nightly MG test automation

[rlratzel]

This PR includes various updates to support nightly automated MG test runs, including:

• Adding a marker which nightly scripts use to run on all visible GPUs instead of a hardcoded number of GPUs, since the scripts rely on knowing the number of GPUs being used in the tests by setting the CUDA_VISIBLE_DEVICES env var.
  • In the nightly scripts, the marker is used like so: pytest -m "not preset_gpu_count" ...
• Added a client.wait_for_workers() call to various setups to both match the approach taken by the MGContext class, and to ensure workers are running. This seemed to increase reliability in the test runs.
  • side note: we should decide to use only the MGContext class or the client_connection pytest fixture in these tests, since they both aim to accomplish the same thing.

[rlratzel]

rerun tests

[codecov-io]

Codecov Report

Merging #1308 (640b3e1) into branch-0.18 (70d9abd) will decrease coverage by 0.16%.
The diff coverage is 38.09%.

@@               Coverage Diff               @@
##           branch-0.18    #1308      +/-   ##
===============================================
- Coverage        60.38%   60.21%   -0.17%     
===============================================
  Files               67       67              
  Lines             3029     3049      +20     
===============================================
+ Hits              1829     1836       +7     
- Misses            1200     1213      +13     

Impacted Files	Coverage Δ
python/cugraph/dask/common/mg_utils.py	37.50% <38.09%> (-2.50%)	⬇️

---

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 70d9abd...640b3e1. Read the comment docs.