map instead of unordered_map for sparse pool

[Yurlungur]

PR Summary

The sparse pool uses an unordered map. However, sparse ids have an implicit ordering, they are integers after all. While not a bug exactly, the unordered map can mangle the ordering and can produce un-intuitive behavior where a user may request sparse IDs and expect the pool to return an ordered list. This PR changes the map to an ordered one (i.e., a heap) so traversal through it always produces sparse IDs sorted from smallest to largest.

PR Checklist

• Code passes cpplint
• New features are documented.
• Adds a test for any bugs fixed. Adds tests for new features.
• Code is formatted
• Changes are summarized in CHANGELOG.md
• CI has been triggered on Darwin for performance regression tests.
• Docs build
• (@lanl.gov employees) Update copyright on changed files

[pdmullen]

LGTM

[lroberts36]

LGTM

[Yurlungur]

Cuda CI test seems to be failing here, but it's working when I run on a local cuda machine. @lroberts36 is reporting similar issues. @pgrete is there perhaps something wrong with the CI machine?

[pgrete]

Cuda CI test seems to be failing here, but it's working when I run on a local cuda machine. @lroberts36 is reporting similar issues. @pgrete is there perhaps something wrong with the CI machine?

Cuda tests did eventually pass, didn't they? And yes, as I mention on Matrix yesterday the machine is more heavily used. now so it's likely just a fight for resources.
Given that it's not "my" node, there's currently very little I can do to fix this.

[Yurlungur]

Cuda tests did eventually pass, didn't they?

Yes eventually. Perhaps just needs to be timed for nighttime in the EU