Add pre-solve scaling support

[FabianHofmann]

Summary

• Add optional pre-solve scaling of constraints and continuous variables to improve numerical robustness
• Row scaling (default) rescales constraint rows by their largest (or RMS) coefficient
• Column scaling (optional) rescales continuous variables so column norms are close to 1
• Scaling is undone automatically on primal/dual values and the objective before storing results
• Integer/binary variables remain unscaled by default to preserve integrality

API

Enable via Model.solve(scale=...):

• scale=True: Enable row scaling with max-norm (default behavior)
• scale="row-max" or scale="row-l2": Select norm for row scaling
• scale=ScaleOptions(...): Full control over row/column scaling, integer handling, etc.

Test plan

• Unit tests for row scaling and dual unscaling
• Unit tests for column scaling with continuous-only variables
• Integration tests comparing scaled vs unscaled solutions
• Benchmark script for performance testing