New features and enhancements
- PyMedPhys now includes its own custom, `numba`-accelerated implementation of
multilinear interpolation. You can find the technical reference
[here](https://docs.pymedphys.com/lib/ref/interp.html).
This was implemented for the following reasons:
- The PyMedPhys implementation gives a 5-8x speed boost over EconForge's
`interplation` and 10-70x over Scipy's `RegularGridInterpolator`. See
the technical reference page for a performance comparison.
- At the time of writing, EconForge's `interpolation` appears incompatible
with `numba>=0.60`.
- Removing an additional dependency (namely, EconForge's `interpolation`)
is generally better for maintenance, holding all else equal.
(Potentially) breaking changes.
- **WARNING**: Updating to `scipy >= 1.11` introduced changes to test
baselines for the electron factors algorithm. This occurred without any
`pymedphys` code changes. Changes in baseline on the order of ~0.5% have
been observed to be, but we cannot guarantee that changes in a given
electron factor calculation do not exceed 0.5%. If you do use this
algorithm, we strongly recommend that you compare current results with
previously known results and report any large deviations
[here](https://github.com/pymedphys/pymedphys/issues/1858)