SOA spherical tensor VGH

[shivupa]

Proposed changes

Paul pinged me about progress on #5391 so I started looking into this again! Very exciting.

Implements SoaSphericalTensor<T>::evaluateVGH needed for backflow on molecules eventually.

I started trying to make a molecular backflow test (found while trying to make a test our of the Li2 results of arXiv:0801.0518).
This was the first missing piece evaluateVGHGH will come in a separate PR, which is needed for optimizations.
In this we apply the derivative recurrence relations to the gradient to get the Hessian.

The storage was sized only for VGL, but I resized the storage for future VGHGH even though the last elements (10-19) are not used yet. This is consistent with SoaCartesianTensor.

I used Codegen as the cartesian path does even though the code is actually simpler since we can apply the recurrence relations.

Sympy checks for the recurrence relations and tests for l<=6 and finite difference tests.

Code generated with GPT 5.6 Sol/Max using the Pi harness and reviewed using Claude Opus 5/Medium.

As far as I can tell the storage changes affect offload.

What type(s) of changes does this code introduce?

• New feature
• Testing changes (e.g. new unit/integration/performance tests)

Does this introduce a breaking change?

• No

What systems has this change been tested on?

Ryzen 9950x workstation, Linux, GCC

Checklist

• • I have read the pull request guidance and develop docs
• • This PR is up to date with the current state of 'develop'
• • Code added or changed in the PR has been clang-formatted
• • This PR adds tests to cover any new code, or to catch a bug that is being fixed
• • Documentation has been added (if appropriate)

[ye-luo]

Test this please

[ye-luo]

Could you move all the function definition currently already outside the SoaSphericalTensor class to the cpp file? We don't need to keep them in the header file.