feat(diff-views): per-bar histogram colors, sparkline rendering, color_static parity

[paddymul]

Renderer and type changes needed by diff views (surfaced by the tallyman diff-view prototyping; the sparkline fixes currently live downstream as CSS overrides on the bundled grid, which this PR obsoletes).

What

• HistogramCell — a HistogramBar may carry an optional color; the population Bar renders one recharts Cell per datum so the color lands on that individual bar. Bars without a color keep the scheme-default fill. This lets a diff view ship a change-distribution histogram whose bars share the exact color key of the cells it summarizes (colors ride the data, so it works in pinned rows where cellRendererParams are not plumbed).
• ChartCell — every line series draws at strokeWidth={1} with dot={false} (the 100×24 chart is CSS-scaled to fill its grid cell, which turns the recharts defaults into a fat blobby squiggle), and a hidden YAxis with 2px top/bottom padding keeps a series whose values sit exactly at dataMin/dataMax — e.g. the collapsed "after" side of a diff — off the plot border where the clip rect swallows half the stroke.
• styling_core — Python catches up to the TS types: DIVERGING_BLUE_WHITE_RED joins ColorMap, ColorStaticRules joins ColorMappingConfig, and ChartDisplayerA grows the colors field the JS already consumes.
• compare.col_join_dfs — join-key columns emit color_static directly, replacing the categorical-map-of-four-identical-colors workaround whose comment ("once the JS is rebuilt") has been stale since color_static landed in the compiled JS.

Why

Diff views color cells by a per-column change key. The summary row needs to tell the same story: a histogram of the change distribution colored by the same key, and before/after sparklines where the flattened "after" series stays visible. At 24px tall, dots plus scaled strokes make lines unreadable, and a floor-hugging series is half-clipped by the SVG border — both currently patched downstream with CSS attribute-selector overrides that reach into recharts internals.

Tests

TDD structure: the first commit carries only the failing tests so CI shows them red; the implementation lands in the follow-up commit on this PR.

• HistogramCell.colors.test.tsx — per-datum Cell fills on the population bar, colored and default
• ChartCell.render.test.tsx — all line series 1px/dotless; hidden padded YAxis present
• compare_test.py::test_join_key_columns_use_color_static — join-key override emits color_static

Deliberately out of scope, noted for follow-up: plumbing displayer_args through getCellRendererSelector for pinned rows (blocks Python-configured chart colors in pinned rows), configurable pinned-row height, and a real per-series color/z-order API to replace the fixed lineRed/lineGray slots.

🤖 Generated with Claude Code

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 99f90b97c7

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[chatgpt-codex-connector]

Keep the new tests green before merging

When this commit is reviewed by itself, this assertion makes the suite fail: col_join_dfs still returns {"color_rule": "color_categorical", ...} for join keys in buckaroo/compare.py, and the same commit also adds JS assertions for ChartCell/HistogramCell behavior that those components still do not render. Because only tests changed here, CI will stay red for every run that includes these tests; include the corresponding implementation in this commit or keep these red tests out of the mergeable change.

Useful? React with 👍 / 👎.