introduce a raster collection concept

[janzandr]

It is proposed to introduce raster collections into the EnMAP-Box data model to manage bigger sets of raster layers (e.g. EnMAP images for a whole country or continent) efficiently. By managing, we refere to:

• filter spatially
• filter temporally
• filter by metadata

Technically, we can just use standard QGIS vector layers for describing raster collection:

• each feature referes to a single raster
• feature geometry is the bounding polygon (or bounding box) of the raster
• required attributes are:
  • raster source
• optional attributes are:
  • start_date, end_date
  • mask_source (external mask associated with the raster)
  • any other type of metadata

Raster collections will be used in the ARD > Aggregate raster layers and other algorithms. Also, we should have an algorithm that creates a Raster Collection by parsing a folder recursively for raster layers.

[janzandr]

@jakimowb, I was also thinking about implementing a custom RasterDataProvider for exploring a raster collection. The provider would be initialized with the raster collection and a desired pixel grid (extent, crs and resolusion). Afterwards the provider can be used for data visualization and reading data with on-the-fly resampling, warping and aggregation (simple mosaicking or more complex aggregations like mean, median etc.).