Benchmark LAO evaluation at point in space

[emielvb]

Current LAO and $\nabla$ LAO evaluations at a point in space (basis.evalBasisSetAtPoint and basis.evalGradBasisSetAtPoint) are not benchmarked for LAO values (they are benchmarked against PySCF for normal GTO basis sets).

In this PR, I aim to benchmark the current implementation.

• ChronusQ: ChronusQ reference data appears hard to get (see https://github.com/GQCG-res/chronusq-reference-data/issues/2#issuecomment-3646093207).
• Grad LAO as finite difference: Finite difference approximation of grad LAO corresponds to current cpp implementation.
• Analytical reference:
  • Current implementation appears well-behaved in reference to analytical behaviour. $k$ can be extracted from evaluated points and its value is as expected.
  • Manual implementation of phase-factor on top of benchmarked bare-GTO point evaluations is in agreement with GQCP's source implementation.