fix: preserve explicit inherited fields

[K1rL3s]

Description

Inheritance mode removed properties explicitly declared by a child schema when they
repeated an inherited property. This discarded child-specific metadata and prevented
downstream profiles from recognizing the declaration.

This PR changes the contract:

• every property explicitly declared by a child remains in the child's IR, even when
  its type, description, constraints, requiredness, and default match the base field;
• compatible overrides render normally — including a $ref narrowed to a schema that
  inherits from the base's (companion: Pet over companion: Creature) and integer
  over number;
• an override the base cannot admit stays faithful to the schema and receives a local
  # type: ignore[assignment, unused-ignore], plus a warning naming the schema and
  property, because a subtype that is not substitutable for its base is a defect worth
  fixing in the spec;
• properties declared only on the base remain inherited and are not copied down.

The marker is stored on IRField.incompatible_override and turned into a suppression
comment in one place shared by the Adaptix, Pydantic, and msgspec renderers. This keeps
generated packages clean under mypy --strict without losing explicit OpenAPI
declarations.

unused-ignore rides along with assignment on purpose: the builder judges
compatibility from the IR, which is a coarser view of a type than mypy's, so wherever
the builder is the stricter of the two a bare [assignment] ignore would be unused —
which --strict reports as an error of its own.

Behaviour change (breaking for consumers)

Keeping a subtype's widening restatement changes decoding, not just typing. Where a
subtype re-declares an inherited property as nullable, the generated model now accepts
null for it, on every serializer:

payload	before	after
LinkButton with text: null	rejected (AggregateLoadError / ValidationError)	accepted, text=None

This matches what the spec declares, and sibling subtypes are unaffected — a subtype
that does not restate the property still requires it. But the annotation consumers see
widens with it: code that read button.text as a str now gets str | None and fails
mypy until it handles the None. Regenerating a client against such a spec is
therefore a breaking change for its callers.

Type of change

• Bug fix (resolves an issue)
• Breaking change (a fix or feature that would cause existing functionality to not
  work as expected) — see the section above
• This change requires a documentation update

Verification

• uv run pytest --cov src --cov-fail-under=80 --cov-report=term - 334 passed,
  100% coverage
• uv run ruff check
• uv run ruff format --check
• uv run mypy
• inheritance compile gate covers Adaptix, Pydantic, and msgspec in both file layouts,
  including the two shapes only the base chain or the numeric tower can justify
• decoding behaviour and the consumer-visible annotation change verified against a
  generated package on both branches

Checklist

• My code adheres to the style guidelines of this project
• I have conducted a self-review of my own code
• I have made the necessary changes to the documentation
• My changes do not generate any new warnings
• I have added tests to validate the effectiveness of my fix
• I have ensured that type checking passes by running uv run mypy
• The generated incompatible override is covered as a concrete code example

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 100.00%. Comparing base (387daab) to head (b2f715c).

Additional details and impacted files

@@            Coverage Diff            @@
##              main       #10   +/-   ##
=========================================
  Coverage   100.00%   100.00%           
=========================================
  Files           31        31           
  Lines         2706      2726   +20     
=========================================
+ Hits          2706      2726   +20     

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.