test(shape-render): cover is_hit_by_point and bounds branches

[wkentaro]

is_hit_by_point is the shape hit-test that canvas selection and hover depend on, yet it had no direct unit coverage despite five distinct branches (line/linestrip edge distance, points never-hit, point radius, mask pixel lookup with an out-of-bounds guard, and the polygon/rectangle path-containment fallback) — it was exercised only transitively through GUI e2e canvas tests. This adds focused unit tests for each branch plus bounds().

The two tests that reach the QPainterPath fallback are left unmarked (no @pytest.mark.gui / qapp): QPainterPath is a Qt value type whose contains() / boundingRect() run with no QApplication, confirmed by running them in isolation.

Test plan

• uv run pytest tests/unit/widgets/_shape_render_test.py -q — 12 passed
• uv run ruff check / ruff format --check / uv run ty check — clean

[wkentaro]

This was generated by AI during PR processing.

Verdict: `recommend-merge`

Test-only PR adding coverage for `is_hit_by_point` and `bounds` in `_widgets/_shape_render.py`. Reviewed the diff against source: the tests exercise every branch of `is_hit_by_point` and their assertions match the implementation.

• line/linestrip -> `nearest_edge_index` (incl. the phantom closing-edge case, asserting the fix(shape): ignore linestrip's unrendered closing edge in hit-testing #2307 fix stays fixed)
• `points` -> never a body hit; `point` -> radius = `point_size / 2` = 4px; empty point guarded to a miss
• `mask` -> translated pixel read with an out-of-bounds guard
• polygon -> path containment; `bounds` -> bounding rect extent

Branch is rebased onto current `main` and CI is green across all 11 configs (lint + 3 OS x Python 3.12/3.13/3.14). No rebase or push needed this tick. Test-only, so no runtime smoke to verify.