thread pool for finding bigrams

[SuperDoxin]

it automatically uses a nice number of processes, or you can specify with --threads.

it doesn't output the same header, but I don't know if this is a regression or if it doesn't matter. haven't the foggiest on how to test for that though.

[Ed-von-Schleck]

Thanks a lot! I'll look at it in greater detail the next days (and try to find out why the result differs), but here's some little things I noticed from glancing over it:

• Try to massage the changes a bit into PEP8 style, with the exception that long code lines are allowed. Specifically, please insert spaces after commas and around operators like = (except for default or keyword arguments).
• multiprocessing.pool() creates a process pool. While there's technically seperate threads involved, it isn't wrong, but a bit misleading to call the argument --threads. Please rename that (and variable names too) to something with processes.

The biggest time consumer is the optimization of the encoding packs, and that's what would profit most of multiprocessing, so if you like, you might want to tackle that, too – that would really be awesome.

Again, thanks a lot, this is a very welcome contribution!

[SuperDoxin]

I stripped out some ordered dicts with normal dicts since the threads can finish in any order. I hoped it wouldn't matter. turns out it might. I'd like to know if it's a problem before I go finish the code up some more.

I'm a terrible terrible person for not following pep8 ;_;