L2 preprocessing

[mdsumner]

get a rough bbox

This for northernmost_latitude etc, files should have this or similar (and how can we get this anyway, surely someone has indexed this for ocean colour?)

ncmeta::nc_atts(f[1]) |> dplyr::filter(variable == "NC_GLOBAL", stringr::str_detect(name, "most")) 

get the boundary in longlat

This is the top,right,bot,left rows and columns in the longlat array as a path.

create VRT

vapour_vrt(sds_var, geolocation = c(sds_lon, sds_lat))

Now, process the VRTs based on the bbox or point in polygon above.

add this logic to GDAL

GDALwarp of VRT should first cull on that bbox, then by the boundary if no bbox, and finally the geolocation arrays should be detected by the driver, so it could potentially be as fast as us doing a pre-filter on the VRTs.

[mdsumner]

the gdal cull might already be done here OSGeo/gdal#7243