Iteratively increase max_distance in radar_dataset

[bart1]

For reading the data it would be good to not omit points of interest when the min_neighbours is not reached. @lyashevska and me discussed that a reasonable procedure would be to increase max_distance iteratively until min_neighbours is reached. The ultimate max_distance should probably be 10 km as we have already omitted data beyond this.

[abelsiqueira]

This is not clear to me yet. Maybe after #66 we can discuss it further.

[bart1]

We discussed the way graphs were calculated. From this line

bird-cloud-gnn/bird_cloud_gnn/radar_dataset.py

Line 29 in 7c166b2

min_neighbours (int): If a point has less than this amount of neighbours, it is ignored.

I get that if annotated points are in sparse regions they are more likely to be ignored. I think that is undesirable, rather it might be good to increase the max_distance for those points so the number of neighbors might be reached

[bart1]

If it easier we can also discuss this later in person

[bart1]

Might also relate to #88

[abelsiqueira]

There is no max_distance anymore, we followed a different approach.