Make program multithreaded and use less memory

[lolpro11]

Use concurrent iterator for cycling through osm entries.
Use rolling buffer to drop data into file when X number of entries have been extracted.

[Tristramg]

Thank you for your interest.
This crate is rather in maintenance mode, so I can’t really say when I’ll have time to spend on this issue, especially given that your need some tricky performance optimization.

The --out is more straightforward. Could you make a separate issue to track it. Do you mean the output format, or the output files (that would require --edges-file and --nodes-file or something in that direction)

[Tristramg]

For the parallelization, with your implementation in #37 we have an ~30% performance improvement. A nice start.

For the memory usage, this is more difficult: we need to parse all the data to count how many times the nodes are used. Unfortunately, the way OSM data in built, there is no situation where we can be sure that partial data is complete and could be dumped