orthonormalize subblocks when multiple species are present

[arthur-lin1027]

No description provided.

[rosecers]

@arthur-lin1027 have you verified this works?

[arthur-lin1027]

@arthur-lin1027 have you verified this works?

This works, but still need to confirm how to contract the two different species, and to see if it can learn the energies well if it is two different species.

[rosecers]

Is this still active?

[arthur-lin1027]

We should keep this active. Let's talk more on if hstacking features is the correct course of action, and hope that energetics are learned with similar performance given two distinct species vs two of the same species, in the same configuration.

[arthur-lin1027]

@rosecers With your approval, I think this can be merged in. It doesn't affect the behavior of single species, and properly handles when there are multiple species, assuming hstacking is the way to go when we have multiple species (which I think it is).