Statmorph

- Change `rms_asymmetry` to `rms_asymmetry2`, since this quantity actually represents the _square_ of A_rms.

- Implement RMS asymmetry (Sazonova et al. 2024).

- More robust search for asymmetry center.
- Attempt Sersic fit even when basic measurements were not successful (flag >= 2).

- Make statmorph slightly more memory efficient.

- An improved initial guess for the double Sersic fit is now obtained by performing single Sersic fits to to the inner and outer regions of the source of interest, which are assumed to be separated by an ellipse of a given size.
- The separating ellipse is by default twice the half-light ellipse, but this can be controlled by the user with the argument `doublesersic_rsep_over_rhalf`
- The Akaike Information Criterion (AIC) and Bayesian Information Criterion (BIC) are now calculated for the single and double Sersic fits.

- Users can now perform _double_ 2D Sersic fits by including the option `include_doublesersic = True`.
- Users can (optionally) customize the fits via the arguments `doublesersic_model_args` and `doublesersic_fitting_args`.
- Furthermore, the option `doublesersic_tied_ellip = True` can be used to ensure that both components have the same ellipticity and position angle, if desired.
- The quality flags `flag_sersic` and `flag_doublesersic` now take values 0-4, just like `flag` (see v0.5.0 release notes).
- Added reduced chi^2 statistics (`sersic_chi2_dof` and `doublesersic_chi2_dof`) to measure goodness of fit.
- Added an example notebook about double Sersic fitting to the documentation.
- Updated tutorial.