feat: add OSQP solver skill

[ConanXu-math]

Summary

• add a bilingual osqp-solver skill for continuous convex quadratic programs
• add a bounded, auditable JSON runner with OSQP 1.x, status, KKT, and infeasibility-certificate gates
• add 27 dedicated regression tests plus a Python 3.10/3.13 GitHub Actions workflow
• update the repository router, English/Chinese documentation, contributor ledger, and optional dependency metadata

Why

OSQP is a widely used QP solver, while this repository did not yet have a dedicated continuous convex-QP workflow. Review of the initial implementation identified false-positive risks in KKT and infeasibility checks plus unsafe CLI boundaries; this version closes those gaps.

Validation

• Python 3.13 + OSQP 1.1.3: python -m unittest discover -s tests -v — 30 passed, 0 skipped
• Python 3.10 targeted package suite during curation preflight — 27 passed, 0 skipped
• GitHub Actions workflow YAML parse — passed
• git diff --check — passed
• local-path, private-key, and token-shape scan — passed

Scope

The included CLI is intentionally a small dense-JSON runner: input up to 32 MiB, n <= 1000, and at most 2,000,000 logical P + A entries. Larger production models should use OSQP's native sparse API.