[x265] [PATCH 0 of 3 ] thread pool and worker thread refactor for NUMA

[Steve Borho]

These three patches will only compile if they are applied together. The slicetype refactor depends on the bonded task groups which depend on the NUMA thread pools, and none will compile without the full set. They are seperated mainly to make the patches easier to review.

Taken together they improve work efficiency on NUMA (multi-socket) machines and allow one to isolate encoders to specific nodes when the use case requires it.

Currently this patch series reduces compression efficiency a bit. This is caused by the introduction of slices in the lookahead, and will likely see further tuning before these patches are pushed.

The new bonded task group structure is very flexible and will allow us to distribute  work in more situations.