Added a new solver for Proximal Averaged Projected Gradient Method

[shnaqvi]

Added a new solver for solving composite prior minimization problem, called Accelerated Proximal Averaged Projected Gradient Method. The optimization problem is given by:

$\text{argmin}_x f(x) + \bar{g}(x),$
$\ \text{where}\ \ \bar{g}(x) = \sum_i \rho_i g_i(x),$
$\ \text{and}\ \ \rho_i \ge 0, \ \sum_i \rho_i = 1$

This addresses the enhancement request #456

[shnaqvi]

Note:

1. New solver has been added as a standalone class deriving from the parent Optimize class. I couldn't make it to derive from the existing PGM class without modifying this other class also, because it combines both the gradient descent and projection steps in one line but for proximal-averaged method, we need to do these steps separately.
2. I couldn't figure out how to use several PGMStepSize routines with this new solver, so I haven't exposed the feature to dynamically evaluate the Lipschitz constant at runtime. It needs to be provided beforehand for now.

[codecov]

Codecov Report

Merging #466 (d694f01) into main (08a5896) will decrease coverage by 0.86%.
The diff coverage is 0.00%.

@@            Coverage Diff             @@
##             main     #466      +/-   ##
==========================================
- Coverage   94.59%   93.72%   -0.86%     
==========================================
  Files          89       90       +1     
  Lines        5541     5592      +51     
==========================================
  Hits         5241     5241              
- Misses        300      351      +51     

Flag	Coverage Δ
unittests	93.72% <0.00%> (-0.86%)	⬇️

Flags with carried forward coverage won't be shown. Click here to find out more.

Files	Coverage Δ
scico/optimize/_papgm.py	0.00% <0.00%> (ø)

[bwohlberg]

Closing PR since alternative solution merged.