Add `Constraint` Pydantic model

[jc-macdonald]

Summary

Add a Pydantic BaseModel for constraint rules, replacing the current ConstraintRule NamedTuple.

Problem

The Pydantic milestone has models planned for every spec sub-structure — Axis (#23), Transition (#21), Mixing (#24), Chain (#26), StateAxes (#27), Operator (#28), ExpressionString (#22) — but not for constraints. ConstraintRule is a NamedTuple with loose types:

• mode: str instead of Literal["allow", "exclude"]
• rules: tuple[dict[str, list[str]], ...] with no validation that rule keys match declared axes

All validation currently lives downstream in ~120 lines of procedural code (_validate_constraint_axes(), _resolve_constraint_mode(), _validate_constraint_rule() in specs.py).

Scope

• Define Constraint Pydantic model with:
  • axes: tuple[str, ...] — validated ≥2 unique entries
  • mode: Literal["allow", "exclude"]
  • rules with validated key→coordinate mappings
• Absorb validation from _validate_constraint_axes(), _resolve_constraint_mode(), _validate_constraint_rule() into model validators
• Update _normalize_constraints() to accept/return Pydantic models
• Update tests

Related

• Constraint filtering for state, alias, and transition expansion #37 (constraint filtering for expansion)
• Constraint-aware sum_over/integrate_over and end-to-end tests #38 (constraint-aware sum_over)
• Document constraints block in README and guides #39 (constraint docs)