Handling of broadening

Putting this paper here, since it may be of interest: https://arxiv.org/abs/2512.09861

A lot of the paper is concerned with how treating line broadening as a convolution is actually an (incorrect) approximation that can matter at high spectral resolution.

In the paper, I also touch on a few things relevant to some issues raised in the past on this repo. My code that wraps Korg is [here](https://github.com/palumbom/FormationTemps.jl). There may be some things that are worth adapting/incorporting into Korg. Specifically:

* https://github.com/ajwheeler/Korg.jl/issues/353#issue-2625296557 -- I implement Gray's radial-tangential macroturbulence as well as the rotation + radial-tangential kernel from [Hirano et al. 2011 Equation B12](https://iopscience.iop.org/article/10.1088/0004-637X/742/2/69)
* https://github.com/ajwheeler/Korg.jl/issues/190 -- You can also get a nice speed-up by handing convolutions on the GPU (depending on how big/fine) your spectrum is. I have GPU methods implemented for the convolutions for micro (which also allows for depth-dependent micro + convective blueshift -- https://github.com/ajwheeler/Korg.jl/issues/431), macro, etc.
* https://github.com/ajwheeler/Korg.jl/issues/332 -- I calculate formation temperatures/heights, optionally including perturbations to the contribution function from rotation & macroturbulence.
* https://github.com/ajwheeler/Korg.jl/issues/273 -- If you broaden by explicitly disk-integrating, rather than convolving, you don't need to use limb-darkening parameters at all. This is much, much slower though.