* Massive reorganization of pointcache code, things are now cleaner than ever.
* For all but smoke the cache is first written to memory when using disk cache and after that written to disk in one operation. This allows less disk operations and the possibility to compress the data before writing it to disk.
* Previously only smoke cache could be compressed, now the same options exist for other physics types too (when using disk cache). For now the default compression option is still "no compression", but if there aren't any problems this can be set to "light compression" as it's actually faster than no compression in most cases since there's less data to write to the disk. Based on quick tests heavy compression can reduce the file size down to 1/3rd of the original size, but is really slow compared to other options, so it should be used only if file size is critical!
* The pointcache code wasn't really 64bit compatible (for disk cache) until now, so this update should fix some crashes on 64bit builds too. Now all integer data that's saved to disk uses 32 bit unsigned integers, so simulations done on 64bit should load fine on 32bit machines and vice versa. (Important disk cache simulations made on 64bit builds should be converted to memory cache in a revision before this commit).
* There are also the beginnings of extradata handling code in pointcache in anticipation of adding the dynamic springs for particle fluids (the springs need to be stored as extradata into point cache).
* Particles were being read from the cache with a slightly wrong framerate. In most cases this probably wasn't noticeable, but none the less the code is now correct in every way.
* Small other fixes here and there & some cosmetic changes to cache panel, but over all there should be no functional changes other than the new disk cache compression options.
* This whole re-organization also seems to fix bug #25436 and hopefully shouldn't introduce any new ones!
* Pointcache code was quite ugly looking and complicated, so here are mostly just cosmetic adjustments, but some improved logic also.
* Slight cleanup of pointcache ui too.
* Shouldn't have any functional changes what so ever, so poke me right away if something seems off.
* Cloth now uses the original derived mesh for simulation and a new dm only for applying the simulation result.
* Reverted Campbell's temporary workaround r33406.
- removed deprecated bitmap arg from IMB_allocImBuf (plugins will need updating).
- mostly tagged UNUSED() since some of these functions look like they may need to have the arguments used later.
* Now cloth reads cache using subframes when rendering.
* Cloth cache also was reset every time on the start frame which kind of defeats the point of the caching.
* Disable openmp for dot product, this gives different results each
time due to non-commutative floating point add.
* Disable openmp with few vertices, the extra thread overhead only
slows things down then.
* Replace the hack that would divide stepsPerFrame and then set it
back, now it simply uses the timescale in the collision function.
This was incorrect because stepsPerFrame is an int, but we don't
want this to be rounded.
* Extra out of bounds check for hair velocity smoothing grid.
to specify different spring lengths.
Implementation is quite ugly because the shape key has to be pulled
through the modifier stack somehow, need a more flexible data mask
system to solve this properly.
(commits 27773,27775,27778 by Brecht from render25 branch)
