Build with -threaded

[pepeiborra]

Otherwise there is no parallelism.
Perhaps consider using setNumCapabilities too.

[xich]

I would totally support this change, but I don't have bandwidth to do it myself right now.

I tried this once (years ago) and had issues with the GHC API crashing. I don't think the GHC session is (or was?) thread safe, or at least not the way retrie uses it (probably the actual problem).

We still get IO concurrency, which is a pretty big win. I suspect most of the time is spent parsing files, but don't have numbers to back that up. (You can run retrie with --single-threaded and see that it is quite a bit slower.) Maybe some sort of benchmark suite with criterion is a good first step?