Reorder atoms in label data (Fixes 502) + documenting C++

[ndickson-nvidia]

Changelogs

• Reorder atoms in node-level and nodepair-level label data, when the same molecule appears in multiple tasks with different atom orders, to be consistent with the first task's atom order. This fixes Re-order atoms and edges when matching molecules from different datasets #502
• Document the important C++ functions

---

Checklist:

• _Was this PR discussed in an issue? Wrong ordering of nodes/edges in multitasking for node/edge level tasks #467
• Add tests to cover the fixed bug(s) or the new introduced feature(s) (if appropriate).
• Update the API documentation is a new function is added, or an existing one is deleted.
• Write concise and explanatory changelogs above.
• If possible, assign one of the following labels to the PR: feature, fix or test (or ask a maintainer to do it for you).

---

discussion related to that PR

[DomInvivo]

I see that canonicalization was used for the re-ordering at dataloading and data preparation time. The only issue is that we can't simply use the "predict mode" for real-world predictions without un-ordering. But I think it's fine. We should probably warn about that in the docs and the code

• Log warning about canonicalization cannot be used for predictions without un-ordering
• Warn in the documentation about the point above