Slow TrueType font atlas generation

[rms7326]

@robertosfield and @AnyOldName3, I posted a vsgXchange pull request #237 to speed up reading of TrueType fonts by performing the glyph contour generation as packaged tasked submitted to a thread pool. This made the font atlas generation much more manageable for our application users who are able to choose any TrueType font on their system during runtime. I am posting this discussion here because I noticed @AnyOldName3's vsgXchange pull request #224 hasn't received any discussion so I didn't know if you routinely monitored vsgXchange for requests. My change mostly preserves your original code, moving a few pieces of logic into lambda's so that they can be packaged up into concurrent tasks with some local state (a glyph quad and the x and y position). @AnyOldName3's changes modify the algorithm and improve the overall contour generation performance. Perhaps the two can be combined in two different merge requests.

[AnyOldName3]

While I've not made anything particularly difficult to parallelise in my PR, doing so won't be as simple as just running the relevant parts of the function at the same time. A significant part of the speedup came from reusing the same intermediate std::vectors for every glyph so most glyphs can be rasterised with no allocations. Obviously, using the same data structures on different threads at the same time will result in problems, so they'll need changing to be per-thread.

[rms7326]

My implementation currently keeps thread local data for the FT_Face, we could easily add thread local data for the std::vector as well. Instead of the entire algorithm reusing the same intermediate std::vector, each thread would reuse the intermediate std::vector from the thread local data - currently called threadFaceData in my implementation because I only needed the single item, the FT_Face, per thread.

[robertosfield]

I've added a comment to @rms7326's PR.

Unfortunately, I have lots of other work on right now that I need to complete before I can get back to reviewing submissions like this so I'll just be able to quick reviews and provide quick feedback, I won't have time to properly dive into topics or test code.