[ephpicman]

Summary

Prepare the WLS-enabling 1.1.0 release candidate on top of the existing OLS foundation.

• Expose Matrix::weightedLeastSquares(Matrix $rightHandSide, array $weights): Matrix using the native implementation.
• Use square-root row scaling plus pivoted QR; do not form weighted normal equations or introduce model-specific fit_wls().
• Document weight semantics, multiple RHS, rank requirements, exception behavior, memory costs, and the Windows support caveat.
• Add a runnable WLS example with deterministic coefficient checks, an end-to-end PHPT, and named-argument/Reflection coverage.
• Update the 1.1.0 changelog, version macro, release roadmap, release-gate audit, documentation index, and source-archive workflow.
• Extend the clean source-archive smoke test to verify version 1.1.0 and run both OLS and WLS examples.

Numerical contract

• Weights must be PHP integers/floats, finite, non-negative, and one per observation.
• Zero weights remove an observation's contribution to the objective; matrix values in those rows are still validated.
• The effective weighted design must have full column rank. Rank-deficient and underdetermined inputs are rejected.
• The implementation materializes square-root-scaled design and response matrices, requiring additional memory proportional to input size.
• Windows remains unverified; this PR does not claim Windows validation.

Validation

All required workflows passed on the exact PR head d54a1e551dcd293fe55aa6ce1ad1dc031736f875:

• Tests: run 571
• Native coverage: run 151
• Source archive smoke test: run 29

The source archive verifies version 1.1.0, runs the PHPT suite, and executes both OLS and WLS examples.