rotation reconstruction convention corrections

[catketchup]

Hi Mat, this is Hongbo. I am using the rotation reconstruction estimators of symlens. It is a very poweful tool.

I checked the implementation of the rotation estimators and found several convention inconsistencies. I notice https://github.com/simonsobs/symlens/blob/master/symlens/qe.py#L1315 which uses the Yadav et. al. 2009 convention. However, I found the norm in https://github.com/simonsobs/symlens/blob/master/symlens/qe.py#L688, the unnormalized estimator in https://github.com/simonsobs/symlens/blob/master/symlens/qe.py#L909 and the N0 in https://github.com/simonsobs/symlens/blob/master/symlens/qe.py#L746 are inconsistent for rotation field reconstruction and don't give the right result. After I corrected the points above in my branch, I compared the in_alpha x rec_alpha with in_alpha x in_alpha, which match well. I also checked the output N0, which matches well with that used in the SO.

Thank you!