Improve performance of stats generation

[ckrause]

We currently iterate over all sequence IDs and check if the program file exists: https://github.com/loda-lang/loda-cpp/blob/main/src/oeis/oeis_manager.cpp#L462

It might be faster to replace this by getting the list of files in the directories and iterating over them.

[RimacC2-EV]

Using multiple LODA instances (-P/-p )to run parallel stats generation might be good?