Introduction of the raft::device_resources_snmg type

[viclafargue]

Introduces the raft::device_resources_snmg type to hold all resources required for the NCCL clique.

Answers #2459
Removed call to raft::comms::build_comms_nccl_only (#2465)

[codecov-commenter]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 81.01%. Comparing base (fccf33e) to head (a269555).
Report is 1 commits behind head on branch-24.12.

❗ There is a different number of reports uploaded between BASE (fccf33e) and HEAD (a269555). Click for more details.

HEAD has 16 uploads less than BASE

Flag	BASE (fccf33e)	HEAD (a269555)
	19	3

Additional details and impacted files

@@               Coverage Diff                @@
##           branch-24.12    #2487      +/-   ##
================================================
- Coverage         87.17%   81.01%   -6.17%     
================================================
  Files                25       17       -8     
  Lines               546      511      -35     
================================================
- Hits                476      414      -62     
- Misses               70       97      +27     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

---

🚨 Try these New Features:

• Flaky Tests Detection - Detect and resolve failed and flaky tests

[cjnolet]

Please add proper documentation to each of the methods in this class (even private methods). It's important for users of this API to know that they'll need to have multiple threads available for this, otherwise it'll hang.

[cjnolet]

Since this is a device_resources instance, I think it'll be better to extend device_resources. That way you can override important methods like get_stream and get_worker_stream to return a main stream. In general, we should expect the device_resources_snmg to behave the same as device_resources when passed into an algorithm that only operates single gpu while being able to operate in snmg mode when passed into an algorithm that supports single-node multi-gpu. I think it should store off and use the id of the GPU that is selected when creating the snmg handle by default when being used in single-gpu mode.