Add professional portfolio PGD experiment

[charlesmartin14]

Summary

• add a production-style projected-gradient portfolio solver with majorization backtracking and live progress callbacks
• validate the quadratic case against exact KKT and SciPy SLSQP
• add nonlinear power-law costs plus realistic factor, sector, box, turnover, and gross-exposure constraints
• add three detailed notebooks, ten automated tests, a Mac /tmp setup script, and a foreground campaign runner with timestamped command-line logging
• save per-run histories, holdings, trades, and CSV/JSON summaries under /tmp/portfolio-pgd-runs
• add path-scoped GitHub Actions validation

Local validation

• python tests/run_tests.py: 10/10 passed
• python scripts/execute_notebooks.py: 3/3 notebooks passed
• logged CLI campaign: quadratic, nonlinear, and realistic scenarios all converged and matched reference solvers within enforced tolerances
• bash -n passed for both Mac shell entry points

Terminal behavior

The campaign runs in the foreground and never backgrounds, detaches, replaces the caller shell, or sends process-killing signals.