improve oshdb-api processing parallelism

[tyrasd]

follow up from #26:

Ideally, one would like to start processing with the lower zoom cells, but have them evenly distributed across all threads in the pool (e.g. in an interleaved way: cell 1 -> thread 1, cell 2 -> thread 2, …, cell n -> thread n, cell n+1 -> thread 1, etc.), but I'm not sure if that's easy to implement using the normal java streams.