Add spark bar chart type for profiler column cards

Problem

Profiler column cards currently display distribution and completeness data as raw numbers, which makes it hard for analysts to visually scan a table with dozens of columns and quickly spot anomalies like heavy skew, missing-data patterns, or uniformly distributed values. Without a compact inline visualization, analysts must mentally compare numeric values across cards — a slow, error-prone process that defeats the purpose of a scannable table overview. A spark-bar mini-chart type tuned for the card's constrained layout would let analysts absorb distribution shape at a glance without expanding into full chart views.

Context

• Profiler cards render spark-bar visuals with stable sizing and theming.
• Spark bars support core states (normal, skewed, missing-heavy) clearly.
• Visual tests cover rendering fidelity in light/dark themes and CI.

Possible Solutions

Plan

• Define spark-bar spec and expected data contract for profiler cards.
• Implement renderer support and template wiring for card integration.
• Add visual regression fixtures for representative column scenarios.
• Tune styling for readability in compact layouts and dark mode.
• Update inspector docs/examples to include spark-bar interpretation guidance.

Implementation Progress

• GitHub issue: https://github.com/fivetran/dataface/issues/282

Review Feedback

• [ ] Review cleared