feat: support tilingMode: cog for raster data type to allow visualization from external stac

[slesaad]

Support external STAC servers via tilingMode: cog

Summary

Adds a new tilingMode option to raster datasets so they can be tiled
either via VEDA's mosaic register endpoint (the existing default) or via
titiler's per-item /cog/tiles endpoint. The latter unlocks consumption of
external STAC servers (Element84 Earth Search, GHG Center, NASA CMR
CloudSTAC, etc.) that don't expose a mosaic register endpoint.

Dataset config

type: raster
tilingMode: cog              # 'mosaic' (default) | 'cog'
stacApiEndpoint: https://earth-search.aws.element84.com/v1
tileApiEndpoint: https://openveda.cloud/api/raster
stacCol: sentinel-2-l2a
searchLimit: 100             # caps items per /search request
sourceParams:
  assets: red
  colormap_name: viridis
  rescale: [0, 10000]

• tilingMode: 'mosaic' (default): unchanged behavior.
  /searches/register produces one tilejson for the whole CQL filter.
• tilingMode: 'cog': per-item COG tiling. For each STAC item, the
  selected asset's href (preferring alternate.s3.href when present)
  is passed to /cog/tiles?url=.... Multiple items render as separate
  layers stacked on the map.

What's also new

• searchLimit: configurable per-dataset cap on items per /search.
  Some external servers (e.g. Element84) error out with limit: 1000.
  Works in both tiling modes.
• Pagination: STAC search responses with a next link now power a
  "Showing N of M items — Load More" overlay. Handles both GET and
  POST next-link methods.
• FootprintsLayer: replaces the marker-based PointsLayer for both
  raster modes. Renders STAC item geometry when the response includes it
  (cog mode), bbox polygons otherwise (mosaic mode). Respects parent
  layer's hidden / opacity / generatorOrder.

File-level changes

File	Change
types/veda.ts, parcel-resolver-veda/index.d.ts	Add tilingMode, searchLimit to DatasetLayer; remove unused external-stac from DatasetLayerType.
style-generators/raster-cog-timeseries.tsx (new)	Per-item COG tiling component.
style-generators/raster-timeseries.tsx	Adds searchLimit + pagination. Mosaic register guarded against re-registration on pagination.
style-generators/footprints-layer.tsx (new)	Polygon footprints layer used by both raster modes.
style-generators/pagination-overlay.tsx (new)	Shared "Load More" UI.
exploration/components/map/layer.tsx	case 'raster' branches on tilingMode === 'cog'.
exploration/hooks/use-stac-metadata-datasets.ts	New branch for type === 'raster' && tilingMode === 'cog' to handle external STAC metadata structures.
mock/datasets/*-external.data.mdx, *-element84.data.mdx	Test datasets exercising the new path against Element84, GHG Center, and CMR CloudSTAC.

Backwards compatibility

tilingMode defaults to 'mosaic'. All existing raster datasets
continue to work without any config change.

Test plan

• Existing internal raster dataset still renders identically (no new
  searchLimit/tilingMode in MDX).
• Mangrove AGB (GHG Center) renders cog tiles when a date with items
  is selected.
• Sentinel-2 / Landsat (Element84) renders cog tiles; "Load More"
  reveals additional footprints without yanking the map.
• HLS L30 (NASA CMR CloudSTAC) renders cog tiles.
• Hide / show / change opacity / change order on a cog layer
  toggles its footprints alongside its tiles.
• Removing a cog dataset from the timeline cleans up its sources
  and layers from the map style.
• Date change while a loadMore is in flight aborts cleanly (no
  console warnings about state-after-unmount).

[netlify]

✅ Deploy Preview for veda-ui ready!

Name	Link
🔨 Latest commit	e36c703
🔍 Latest deploy log	https://app.netlify.com/projects/veda-ui/deploys/69f16cfc6d0f34000896d95e
😎 Deploy Preview	https://deploy-preview-1982--veda-ui.netlify.app
📱 Preview on mobile	Toggle QR Code... Use your smartphone camera to open QR code link.

---

To edit notification comments on pull requests, go to your Netlify project configuration.

[dzole0311]

Suggested change

	const LOG = process.env.NODE_ENV !== 'production' ? true : false;
	const LOG = process.env.NODE_ENV !== 'production';

[dzole0311]

Looks like we could reuse StacLink, PaginationState and the loadMore callback between the two timeseries files