Simplify osrm-extract file output

[danpat]

Edge-based-graph expansion was moved into osrm-extract by simply transposing the code from osrm-prepare.

The code is inefficient, in that it writes the node-based graph to disk, then re-reads it, and performs edge expansion. Now that all these steps happen inside osrm-extract, there are some steps that can just stay in memory, rather than going via intermediate files.

This would save some IO, and remove a few of the output files.

The code that was moved from osrm-prepare starts here:

osrm-backend/src/extractor/extractor.cpp

Line 253 in 54ee76b

{

[daniel-j-h]

Can we table this for 5.1? I see no immediate priority in pushing this forward in the next days.

Let's ship v5.

[TheMarex]

Alright, my main goal for 5.1 is to redo our storage layer anyway.