Inconsistent data fidelity scaling between different algorithms

[andrewwmao]

After seeing Issue #114 on MRIReco.jl, I noticed that certain algorithms (e.g. ADMM) scale the data fidelity term by 1/2, but some do not (e.g. FISTA). This creates an undesirable problem where the same lambda gives different results on different algorithms. We have actually seen this first hand with our own data.
I would be inclined to fix this just for FISTA by multiplying the Lipschitz constant by 2, but I wonder if the package wants to decide on a consistent way to handle this across the board.

[tknopp]

yes, absolutely. If you would have some time correcting these it would be much appreciated.

In my opinion the data fidelity should always be scaled identically. It would be easier to remove the 1/2 for ADMM probably since Kaczmarz for instance also does not have this pre-factor.

[andrewwmao]

I am not familiar enough with the algorithms outside of FISTA & ADMM, but I can make the change to ADMM if that's what will make things more consistent overall.