ADMM, DPGD and ChambollePock don't cv to the min

As we can see in the following figure, although the algorithms of ADMM, Dual proximal GD and Chambolle-Pock converge, their resultats are slightly different from the others. Maybe it is caused by the dual variable or proximal gradient, however, the algo of CondatVu also uses dual variable and proximal gradient, it converges well like others.


Start by reproducing the comparison shown in the linked figure for ADMM, Dual proximal GD, Chambolle-Pock, and Condat-Vu. No source file or test is named; inspect the implementations and convergence results to identify why the first three do not reach the minimum, then verify that their results agree with the expected minimum.

Written by the indexing model from the issue text.