split_job_s2grid fails when original GeoDataFrame already contains a tile column

[VincentVerelst]

The split_job_s2grid function uses the Geopandas sjoin_nearest method on the tile columns in the S2 grid. However, if the original reference data already contains a tile column, the sjoin_nearest method will generate a left_tile and right_tile column.

This will cause the groupby('tile)` method to fail: https://github.com/Open-EO/openeo-gfmap/blob/main/src/openeo_gfmap/manager/job_splitters.py#L95