Investigate fitter and recentering_boxsize parameters in epsf_builder

[emirkmo]

Currently called like so on:

flows/flows/photometry.py

Lines 110 to 118 in 01baac6

	logger.info("Building ePSF...")
	builder = self.epsf_builder(
	oversampling=1, shape=1 * self.star_size,
	fitter=EPSFFitter(fit_boxsize=max(int(np.round(1.5 * self.fwhm)), 5)),
	recentering_boxsize=max(int(np.round(2 * self.fwhm)), 5),
	norm_radius=max(self.fwhm, 5), maxiters=100,
	progress_bar=multiprocessing.parent_process() is None and logger.getEffectiveLevel() <= 20
	)
	epsf, stars = builder(stars)

In practice this means 5x5 maximum, which may be unsuitable for some NIR images!
The recentering_boxsize is also the same.

The maximum was put in place to get infront of bad FWHMS creating a poor EPSF, which is way more common than needing a larger boxsize.

As step 1, this maximum should be based on the typical FWHM per instrument, and instruments with poor seeing or large FWHM or small pixels should have this increased!

Step 2, Investigate whether a max that is less than the star size is even needed.

Step 3, There should be a better way to set the max than statically per instrument. (Perhaps just use historical data from pipeline to update this dynamically), or something else to ensure we have a good FWHM.