[FEA]: Add parameter to prevent persisted edgelists in datasets API

[nv-rliu]

Is this a new feature, an improvement, or a change to existing functionality?

Improvement

How would you describe the priority of this feature request

Low (would be nice)

Please provide a clear description of problem this feature solves

When cugraph.datasets objects are used to clean-up MG tests (ex. #4197), they often need to store edge-lists for SG and MG (dask_cudf) usage. However, the current implementation of datasets requires constant calls to unload to avoid these issues.

This also happened to interfere with CI due to the fact that edge-lists were persisted between files.

Describe your ideal solution

Similar to how MG algorithms have a flag that developers use for testing/debugging (perform_expensive_check), perhaps the datasets API should also have a flag that is set when used for testing purposes in order to automatically check for preexisting edge-lists and unload them.

from cugraph.datasets import karate
df = karate.get_edgelist()
ddf = karate.get_dask_edgelist() # This just returns a cudf.DataFrame instead of dask_cudf

# proposed solution
df = karate.get_edgelist(auto_unload=True) # prevents edge-list from persisting for test usage
ddf = karate.get_dask_edgelist(auto_unload=True)

Describe any alternatives you have considered

Since this issue only affects tests, an alternative could be to use fixtures that perform the "check and unload" steps in each unit test.

Additional context

This is part of a general effort to improve readability of the MG tests #4187

Code of Conduct

• I agree to follow cuGraph's Code of Conduct
• I have searched the open feature requests and have found no duplicates for this feature request