[WIP] Behaviour with Korg.Fit.ews_to_stellar_parameters when no EW errors given

[andycasey]

The documentation regarding measured_EWs for Korg.Fit.ews_to_stellar_parameters says that if no EW errors are specified, all lines are assumed to have the same uncertainty. This is kinda true but leads to some unusual behaviour.

All lines are assumed to have the same uncertainty in EW of 1 mA (measured_EW_err=ones(length(measured_EWs))).

This sets equivalent uncertainties in EW, which is not the same as equivalent uncertainties in abundance, because strong lines (e.g., 200 +/- mA) end up getting weighted more than weak lines (e.g., 20 +/- 1 mA).

This is a WIP for now because I think Adam would prefer a default_EW_error function to set the default value.

[ajwheeler]

Do you still believe in this given what we now know about EW fitting?

[andycasey]

I do because the estimated error on EW is usually fractional (a percentage), but I'd like to do some science tests with the revised EW methods before strictly proposing how EWs are handled when no uncertainties are given.

If you are keen to do housekeeping on GitHub issues then you can close this PR and add it to an item in the roadmap/TODO list.

[ajwheeler]

Happy to leave it open as long as we still want to investigate.