fix(chart): audit and repair the chart component across its 12 types

[LeadcodeDev]

What happened

While building examples/effect-ts.json I needed a chart comparing native TypeScript against Effect. I used horizontal_bar with two points, one of them zero, and the render was unusable — a flat light band with no axis, no labels and no second bar. I replaced it with two number cards rather than ship something visibly broken, so the defect is still there.

What is already known

• chart was one of the 23 components with no size source at all; fix(components): give the 23 unsized components a size source #126 gave it a default (400×300 landscape, 320×320 for pie/donut/radar/radial_bar). So the sizing half is fixed, and the remaining problem is in the painter.
• The audit's card matrix classified chart as "renders correctly once explicitly sized", but that was a single configuration — chart has 12 types and only one was exercised.
• A zero value is the specific case that broke: a bar of height zero should still render its label and its axis slot.

What this issue asks for

Audit the chart component the way the earlier capability audits were run — measured, not eyeballed:

1. Render every one of the 12 chart_type values (bar, line, pie, donut, horizontal_bar, area, stacked_bar, radar, scatter, radial_bar, funnel, waterfall) with realistic data, both standalone and inside a card. Report which render correctly, which render but are mis-laid-out, and which produce nothing.
2. For each, check the parts separately: bars/points, axes, gridlines, labels, legend, value annotations. A chart that draws its series but no labels is unusable in a video and should be reported as such.
3. Cover the degenerate inputs a real author hits: a zero value, a single data point, all values equal, negative values, more categories than fit the width, long category labels.
4. Establish whether the legibility floor added in fix(validate): check content against its card, not only against the frame #128 covers chart text — the audit noted chart labels as explicitly not covered.

Every visual claim proven by rendering and measuring pixels. A passing unit test is not evidence a chart is readable.

Acceptance

• A matrix of the 12 types × the parts above, with the evidence.
• The horizontal_bar case from effect-ts.json renders as a usable comparison chart, labels included.
• Degenerate inputs either render sensibly or fail validation with an actionable message — not silently.
• Documentation in .claude/skills/rustmotion/rules/chart-types.md corrected against what was measured.