1D missing data WITH missing filter.

The descent seems stable and faster than steepest descent.
It does not seem to become quadratic convergence.
Is it proper CG and is CG inferior to partan?

It uses -lbook3 routine cgstep

--------------------------------------------------------------
To Do
Find out how to make the method ``standard'' CG or partan

scale input signal and filter to be commensurate

damped filters and prior filters.
