One-body density matrix test failures in batched real code but not complex

[prckent]

Describe the bug

I noticed the short tests are currently reliably failing for real CPU builds. Many of these checks are "checksums", so there are at least normalization issues to fix. I also noticed that some of the test tolerances report as Inf. These tests are not regularly run on GPUs so we don't know the cdash status for them, but presumably they fail for the same reasons.

Recent cdash failures

To Reproduce

Nightly, develop

Expected behavior

Test passes

System:

nitrogen, nitrogen2, sulfur.