A rework for the advection module that (should) improve WENO performance

[simone-silvestri]

This PR reworks how advection is implemented by passing a reduced_order integer to the interpolate functions.

At the moment, upwind and centered reconstruction do not formally change; the ifelse that was upstream before in the alt_interpolate functions is just pushed downstream in the reconstruction function.

For WENO, the reconstruction approach changes because the order is accounted for by changing the reconstruction coefficients and the smoothness coefficients.
This method should give us a boost in performance in the WENO case for bounded and immersed boundary grids, as there is much less computation to be performed. I will post some benchmarking later on.

This PR is still exploratory, so there is a bit of benchmarking and cleaning up to do.

[simone-silvestri]

Results of benchmarking:

• The diff_periodic_output.txt which is marginally affected by this PR (in the vertical direction only), shows and improvement of around 30% for tracer kernels and 10% for velocity kernels

• To confirm this, the periodic_cheap_advection_output.txt which should not be affected at all by this PR, ideed shows ony a very mild improvement for both tracers and velocity kernels (about 3%) which is comparable to noise (altough is consistent between kernels)

• The diff_bounded_output.txt which should be affected by this PR, shows an improvement of about 60% for tracer kernels and 20% for velocity kernels.

• The diff_immersed_output.txt which sohuld be very much affected by this PR yields an improvement of about 80% (and more) for the tracer kernels and about 30% for the velocity kernels