Add fictional commitment challenge set for partner discovery

[kpernyer]

Add a small fictional challenge set to test the commitment abstraction before generating training data. Eight cards cover rejected deliveries, cancellation, resource contention, late evidence, intervention after a forecast, disputed meaning, routine fulfilment, and work that may not fit a bounded commitment.

Each card separates decision-time knowledge from later observations and keeps unresolved interpretations explicit. A contract-question map, reusable case template, and partner discovery guide help practitioners introduce their own distinctions before seeing the project's examples. The README and initial work plan link to the material.

These are discussion scenarios, not labelled training examples or partner findings. The 0.1.0 schema, validator, and executable fixture are unchanged.

Validation: checked local documentation links, balanced disclosure blocks, eight fictional case declarations, and git diff --check. No runtime behaviour changed.