Modularize Libint2 constructor

[susilehtola]

Description

To implement the SAP guess (#3036), I need to toy with the contraction coefficients. Currently, there is duplicated data in BasisSet, as the Libint2 shells are built alongside the basis set itself. The idea of this PR is to modularize the Libint2 shell data creation so that the data can be refreshed based on the data in the BasisSet.

The PR also fixes an issue in that the n_prim_per_shell_ member of BasisSet was never populated.

User API & Changelog headlines

• RN 1
• RN 2

Dev notes & details

• Feature1
• Feature2

Questions

• Question1

Checklist

• Tests added for any new features
• All or relevant fraction of full tests run

Status

• Ready for review
• Ready for merge

[JonathonMisiewicz]

Let me make sure I understand this:

The old behavior was that l2_shells_ was being populated at object construct time. This makes perfect sense if various fields used in constructing the libint2::shell were const, but they are not. Instead, we're migrating the libint2 construction to its own function so it can be called if these fields ever get changed, as #3036 does.

Is that all correct?

[susilehtola]

Let me make sure I understand this:

The old behavior was that l2_shells_ was being populated at object construct time. This makes perfect sense if various fields used in constructing the libint2::shell were const, but they are not. Instead, we're migrating the libint2 construction to its own function so it can be called if these fields ever get changed, as #3036 does.

Is that all correct?

Yes. The point is that if you change data in BasisSet, the data in the Libint2 shells don't update, and one needs the ability to refresh that data as well based on the master data in BasisSet.

[JonathonMisiewicz]

Looks good to me. I ping @jturney on libmints changes, by default.

[jturney]

Seems good to me!