fix: polygon filtered index per layer, instead of per dataset

[igorDykhta]

For #3277

Polygon filters were applied at the dataset level, modifying a shared filteredIndex that ALL layers on that dataset use. When two layers (e.g., "pickup" and "dropoff") share the same dataset:

• Single-layer selection: Filtering by one layer's positions removed rows from the dataset, causing the other layer to lose data points even though it wasn't targeted.
• Multiple layers selected: The filter used AND logic (.every()), requiring a row's position to be inside the polygon according to ALL selected layers simultaneously — which rarely happens when layers use different coordinate columns.

Polygon filtering was moved from the dataset level to a per-layer level:

• Polygon filters were excluded from the dataset's CPU filter pipeline (getFilterRecord), so dataset.filteredIndex is no longer affected by them.
• A new filteredIndexByLayer property was added to each dataset, computed by a new computePolygonFilteredIndexByLayer function that applies polygon filters independently per-layer using each layer's own position accessor.
• Each layer now uses its own polygon-filtered index (from filteredIndexByLayer[layerId]) when building its render data, falling back to the base filteredIndex if it's not targeted by any polygon filter.
• This ensures that polygon filters only affect the layers they explicitly target, leaving other layers on the same dataset untouched.

[Copilot]

Pull request overview

This PR fixes polygon filtering behavior so that polygon filters are applied per-layer rather than at the shared dataset level, preventing unintended cross-layer filtering when multiple layers share a dataset (and avoiding the prior “AND across layers” behavior when multiple layers are selected).

Changes:

• Excludes polygon filters from the dataset-level CPU filter pipeline so dataset.filteredIndex is no longer impacted by polygon filtering.
• Adds filteredIndexByLayer to datasets and computes per-layer polygon-filtered indices in KeplerTable.filterTable.
• Updates layer data/update triggers to use the per-layer filtered index when present, and adjusts unit tests/helpers accordingly.

Reviewed changes

Copilot reviewed 6 out of 6 changed files in this pull request and generated 1 comment.

Show a summary per file

File	Description
test/node/reducers/vis-state-test.js	Updates polygon filter tests to assert per-layer filtering and unchanged dataset filteredIndex.
test/helpers/comparison-utils.js	Adjusts dataset comparisons to optionally ignore the new internal filteredIndexByLayer field.
src/utils/src/filter-utils.ts	Removes polygon filters from dataset CPU filtering via getFilterRecord while keeping them tracked in the record.
src/table/src/kepler-table.ts	Introduces filteredIndexByLayer and computes per-layer polygon-filtered indices during dataset filtering.
src/layers/src/mapboxgl-layer.ts	Uses per-layer filtered index for update triggers / data building in Mapbox GL layer.
src/layers/src/base-layer.ts	Uses per-layer filtered index when updating meta and calculating data attributes for a layer.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.