Safer default facetselfcal config for delay calibration
For complicated sources the current default settings for facetselfcal are sub-optimal. I've almost never not used it, so I think it is safe to always go for the "reset" approach where phases are first solved on the international stations and the remote stations are subsequently solved for on longer timescales or with a higher smoothness constraint.
The screenshots below compare the 9th iteration on a couple of LBCS candidates between the current approach on the left and the reset approach on the right, all starting from identical point source models. It seems to consistently match or outperform the current default.
**Default config**
```
imsize = 1600
pixelscale = 0.075
robust = -0.5
uvmin = 40000
maskthreshold = [7.0]
soltype_list = ['scalarphasediff','scalarphase','scalarcomplexgain']
soltypecycles_list = [0,0,2]
solint_list = [4,1,100]
nchan_list = [1,1,1]
smoothnessconstraint_list = [10.0,2.0,10.0]
antennaconstraint_list = ['alldutch',None,None]
docircular = True
forwidefield = True
stop = 10
beamcor = "auto"
phaseupstations = "core"
```
**Suggested "reset" config**
```
imsize = 1600
pixelscale = 0.075
robust = -0.5
uvmin = 40000
maskthreshold = [7.0]
soltype_list = ['scalarphasediff','scalarphase','scalarphase','scalarcomplexgain']
soltypecycles_list = [0,0,0,2]
solint_list = [4,1,1,100]
nchan_list = [1,1,1,1,]
smoothnessconstraint_list = [10.0,2.0,10.0,10.0]
antennaconstraint_list = ['alldutch',None,None,None]
resetsols_list = [None,'alldutch',None,None]
docircular = True
forwidefield = True
stop = 10
beamcor = "auto"
phaseupstations = "core"
```
I can open a MR for further discussion or testing.
issue