CLI/API consistency and docs

[tomalrussell]

General notes on high-level API and command-line interface consistency:

• file format - allow either geo (geopandas.read_file,to_file) or parquet both input and output
• snail split and snail process should consistently specify output file name
• both should be able to handle more than one band (or document limitation)
• handle reading specific layers from vector file
• add examples to README and docs
• document that snail process calculates all features intersected with all rasters - one features CSV row creates one output file with columns for each raster file/band
• attribute column names in output
• split_linestrings should check for MultiLineString and either throw error or coerce (df.explode)
• check where CRS assumptions must hold! e.g. it should be okay to overlay_raster at the top level and implicitly do reprojections if vector/raster CRSs don't match.
• apply_indices should be done inside split_linestrings
• name high-level functions overlay_raster[s] to take (features, raster dataset) and return (split features with attributed data)
• avoid need for output_path column - if not present, name it like "{feature_filename}_{feature_layer}__{rasters_filename}"
• default (or option?) to processing all layers in a layered vector file