Data inspector: input/resolved rows below the chart, with a resizable divider

A collapsible Data panel under the Live Preview and Chart Builder charts shows the rows the chart actually uses, switching between Input (parsed source) and Resolved (post-transform) views read from the live Vega view. Collapsed by default; the open-state and a draggable height divider persist.

Rows come through a new RenderHandle.inspectData() accessor, so no component touches the view: core/result-data picks the most-upstream source and most-downstream result from the compiled dataflow, read lazily. The divider reuses the window-splitter pattern (horizontal variant).

Consolidations: a shared DataTable primitive replaces the inspector's and the builder's duplicate read-only tables; useResizeDrag merges the col/row drag-gesture twins.

Docs: spec 04/06 and arch 05/10 updated; the now-shipped exploration memo removed.

docs/spec/04-live-preview.md

@@ -71,6 +71,18 @@ Layout: a saved-theme list on the left; the open theme's editor on the right.
 
 The preview pane header also carries a per-chart **Export** control — a disclosure for copying or downloading the current chart's spec, or downloading its rendered image (PNG/SVG). It exports what the preview shows. The behavior is specified in _Import & Export → Per-chart export_; it lives in this header because the image formats are produced from the live rendered view.
 
+## Data Inspector
+
+Below the chart, a collapsible **Data** panel shows the rows behind the chart — the answer to "why is my chart empty or wrong" is to look at the data the spec produced. It reads the live rendered view, so it reflects exactly what was drawn.
+
+- A **disclosure** at the bottom of the pane, collapsed by default (a debugging aid, not the default view); the open state persists across sessions, stored in _Settings_ as `ui.dataInspectorOpen`.
+- An **Input | Resolved** switch chooses which rows to show:
+  - **Resolved** (default) — the rows the chart draws, _after_ the spec's transforms (filters, calculated fields, aggregation).
+  - **Input** — the parsed source rows, _before_ those transforms. A spec with no transforms shows the same rows for both.
+- The table is read-only and capped (the first rows, with a "first N of M" note), like the dataset and builder previews, and updates whenever the chart re-renders.
+- States: when nothing has rendered, the panel guides the user to render a chart; when the chosen view's table is empty, it names which side is empty and why (the resolved side's transforms produced nothing to draw; the input side's source has no rows).
+- A **draggable divider** between the chart and the panel sizes the panel's height (the chart above absorbs the change) — the same window-splitter interaction as the layout panes (see _Application Shell & Navigation_). The height persists, stored in _Settings_ as `ui.dataInspectorHeight`.
+
 ## Rendering Contract
 
 Before the chart is drawn, the spec shown in the editor is transformed into the spec actually rendered. Two deterministic transforms are applied in order. They are specified here because reproducing them faithfully is what makes references and fit modes behave correctly; the result is observable as the rendered chart.

docs/spec/06-chart-builder.md

@@ -52,7 +52,7 @@ The section is ordered **input → shaping** so the distinction reads at a glanc
 ### Data preview
 
 - The first item in the section: a collapsible, **read-only** sample of the dataset's first rows (capped), with a per-column **type chip** in each header. It lets the user sanity-check inferred types against the actual values _before_ building — exactly when type inference is most likely to surprise. Editing the data is out of scope. A non-tabular payload (a single JSON object, TopoJSON, an unfetched URL) has no rows to show.
-- The preview shows the dataset's **raw source rows** — _before_ the filters and calculated fields below are applied; it does not show derived columns. Its position above those controls makes that explicit: it is the input, not the result. (The transformed result is what the right-hand chart renders. A transform-aware "data inspector" that shows the resolved rows is a planned future direction — see [`docs/exploration/data-inspector-exploration.md`](../exploration/data-inspector-exploration.md).)
+- The preview shows the dataset's **raw source rows** — _before_ the filters and calculated fields below are applied; it does not show derived columns. Its position above those controls makes that explicit: it is the input, not the result. The transformed result is shown by the **data inspector** under the chart on the preview side (below), which switches between input and resolved rows read from the rendered view.
 
 ### Filters
 
@@ -147,6 +147,7 @@ A slim strip pinned **under the live preview** — these describe the chart itse
 - Updates are debounced: changes to mark, encodings, or dimensions trigger a re-render after a short pause rather than on every keystroke.
 - While no encoding is mapped, the pane shows a placeholder instructing the user to configure at least one encoding.
 - If the spec fails to render, the pane shows an inline error message describing the problem instead of a chart.
+- Under the chart, the same **data inspector** as _Live Preview_ (Input | Resolved rows, collapsed by default) lets the user compare the source rows against what the builder's filters and calculated fields produced. It appears only with a live chart, so it never doubles the placeholder/error.
 
 ## Validation