Performance bottleneck identified in lower incomplete gamma function

[hatemhelal]

@AlexanderMath identified that much of the time in evaluating eri_basis_sparse is spent evaluating the lower incomplete gamma function here:

https://github.com/graphcore-research/pyscf-ipu/blob/023e11487762731117d73d14bc98b2f74226f800/pyscf_ipu/experimental/special.py#L95