JOSS review: Paper edits

[arjunsavel]

Describe the bug
There are a few small edits to the manuscript for this project's corresponding JOSS paper that I'd suggest.

• like brown dwarfs and exoplanets —> such as brown dwarfs and exoplanets
• fully-polarized —> fully polarized
• flexible and —> flexible, and
• orders-of-magnitude —> orders of magnitude
• barrier-of-entry —> barrier of entry
• redesigns the radiative transfer code —> designs a radiative transfer code
• like GPU —> such as GPU
• Perhaps describe why GPU acceleration and autodiff are beneficial for retrievals
• high fidelity —> high-fidelity
• Was RT defined?
• 0.01 step size —> 0.01 cm^-1 step size
• Are spectra obtained from ExoMOL and HITEMP, or are the opacity data obtained from them, using HITRAN by default otherwise?
• The Q and U benchmark comparisons should also be stated in percent, for consistency (and to make it clear that it's a fractional comparison).
• Define auto-differentiation
• Like delta truncation —> such as delta truncation
• Note that the 100x speedup is only for one of the simulations (10^5 spectral points), as opposed to lower numbers of spectral points.
• performance optimizations in vSmartMOM.jl suggest —> performance optimizations in vSmartMOM.jl, suggest
• Could the comparison to literature tables be presented as a plot, as well? It's a bit tricky to absorb in a table.
• Nice benchmark to other codes! It’d also be useful to briefly describe other codes and differences. e.g., do all radiative transfer codes use line-by-line calculations?

[arjunsavel]

Thanks for making these changes! The paper looks excellent!