[codex] add distribution chart intents

[tmusser]

Summary

This PR expands chart-contract beyond trend, rank, and compare charts with first-class distribution intents:

• Chart.histogram()
• Chart.boxplot()
• Chart.violin()

It also adds distribution-specific audit rules so these charts do not just render - they declare and check the sample-size, grouping, unit, binning, and density-shape assumptions behind the claim.

Why

Distribution charts are where beautiful visuals can quietly smuggle weak claims.

Histograms can imply shape from arbitrary bins. Boxplots can hide thin group sizes. Violin plots can make low-sample data look more precise than it is. This PR keeps those visual forms aligned with the repo's core promise: claim-first, audited analytical charts.

What changed

• Added public distribution chart constructors.
• Added Altair renderers for histogram, boxplot, and violin intents.
• Added distribution audit rules for numeric value fields, sample size, grouped sample size, histogram bins, and violin sample-size risk.
• Added deterministic distribution chart examples.
• Updated README, audit rule docs, roadmap, and example outputs.

Validation

• python -m pytest -q
• python examples/bad_to_good_chart.py
• python examples/distribution_charts.py
• python -m py_compile src/chart_contract/*.py src/chart_contract/renderers/*.py
• git diff --check