Improve solution for #218

[SwooshyCueb]

The current solution for #218, while fine, could use some improvement. It would be nice to migrate from a thread state based approach to a GIL state based approach, and to implement temporary/early unlocking of the GIL.

[korydraughn]

What are the benefits of moving to a GIL state based approach?
Are there any downsides?

[SwooshyCueb]

I've done a lot of timing tests and a GIL state based approach is significantly faster when it comes to nested locking, especially in older versions of python, but has more or less the same performance in other scenarios. In my testing I haven't found any downsides (yet).

[korydraughn]

Sounds good.