Cvxpnpl Versions Save

This release:

• Adds support to SCS 3.x.x
• Add type annotations to the public interface

Fixed issue with requirements file parsing specific to pypi

Fix version parsing issue in setup.py

• Optimized the pose composition after solution candidates are returned.
• Add check for duality gap to validate the global optimality of the solution.

Auxiliary Code:

• Massively refactored the benchmarks.
• Added real data benchmarks.
• For more info check the benchmarks page.

Stripped the intermediate layer previously mediated by CVXPY and now we're directly invoking SCS. Also bumped minimum version of SCS to 2.0.0, which witnessed an increase in convergence performance.