Behavioral Consistency and NumPy Dispatch Compatibility Hardening

[francof2a]

Title

v0.5.0: Behavioral Consistency and NumPy Dispatch Compatibility Hardening

Problem Statement

We completed most v0.5.0 stability work, but key behavioral-consistency items are still open. We need to close remaining correctness and compatibility gaps before moving deeper into feature/performance work.

Why This Matters

• Reduces risk of subtle regressions in equality checks and dispatch behavior.
• Aligns runtime behavior, tests, and docs for op_input_size defaults.
• Closes the remaining warning-path gap in NumPy casting compatibility.

Scope

• Confirm and fix potential Fxp.equal index handling bug (index is None semantics).
• Add tests for Fxp.equal with index=None, scalar index, and slice index.
• Reconcile op_input_size default behavior between runtime and documentation.
• Add explicit tests that lock expected default op_input_size behavior.
• Review all dispatch-implemented functions for NumPy 2.x argument compatibility.
• Add compatibility tests for each currently implemented dispatch decorator target.
• Resolve ComplexWarning in fxpmath/utils/numeric.py cast path, or document+test the expected warning behavior.

Acceptance Criteria

• Full relevant test suite passes on NumPy 1.26.4 and latest 2.x.
• Fxp.equal index behavior is explicit, correct, and test-covered.
• op_input_size default behavior is consistent in code, tests, and docs.
• Dispatch kwargs behavior is consistent across implemented targets on supported NumPy versions.
• ComplexWarning path is fixed or intentionally documented with regression tests.

Deliverables

• Runtime/code changes for behavioral fixes.
• New/updated regression tests.
• Documentation updates for any user-visible behavior clarifications.

Out of Scope

• New dispatch features (vecdot, matvec, vecmat, norms).
• Performance optimization work not required to satisfy correctness/compatibility.