Optimize Pippenger so that it uses num_initial_points = size of the span rather than start_index + size of span

[lucasxia01]

In AztecProtocol/aztec-packages#9098, we set the number of points to process in pippenger to be equal to the start_index + the size of the scalars span. However, it should be pretty easy to only look at the scalars span instead, which would optimize commitments slightly.

This issue was introduced due to the above PR, and not because of any prior PRs.