feat(dashboard): rebuild the § 02 quality breakdown tables

[contactacefleetdev]

The engine/model and task-domain tables carried every style inline and read as two flat dumps: engines and their underlying models sat in one undifferentiated list, and eleven columns of bare figures gave no way to rank a row without parsing digits.

• group rows split agent engines from underlying models inside one ruler of columns, so the two populations never read as one list
• completion, verification, and first-pass each gain a 2px meter under the figure, colored on the same thresholds the stat tiles use
• score chips take a fixed width; the column now scans as a straight edge
• section titles become mono eyebrows with a hairline rule and a count, and column headers drop to the page's standard 9.5px mono from 12px bold
• zeros and unit suffixes recede to --ink-4 so real values carry the contrast
• domains sort by session volume rather than taxonomy order, and both tables scroll inside their own wrapper instead of widening the page

The per-cell inline styles move to a .qt/.gr/.mt/.eng block and each row is built from four helpers, which is where most of the line count went.

Description

Summarize the changes made in this Pull Request and the problem being solved.

Type of Change

• Bug fix (non-breaking change which fixes an issue)
• New feature (non-breaking change which adds functionality)
• Breaking change (fix or feature that would cause existing functionality to not work as expected)
• Documentation update

Verification

Describe the tests and checks performed to verify changes:

• pytest passes cleanly.
• python -m build and twine check dist/* pass.
• E2E test script scripts/e2e_test.py passes.

Checklist

• My code follows the code style of this project.
• I have updated documentation where necessary.
• I have updated CHANGELOG.md with my changes.