Found and fixed a bug I would've never noticed in abelian sandpile:
the png is technically rotated by 90 using the standard WIDTH then
HEIGHT for loop, so I had to switch them. As the abelian sandpile
fractal is 4 way symmetric this wasn't noticeable.
Less intense on the main thread now, no memory leaks so far.
To ensure we're not wasting cycles on updating an already completed
texture, I use both threads_done and rendered which allows for
checking if a new render batch has been started or not. This allows
us to only update the texture if rendering has not completed.
Essentially using a naive threaded solution with a naive escape
routine for choosing whether a given location is actually in the
Mandelbrot set. I think the colouring solution is quite unique: I use
a scaled quadratic function which ensures maximum colouring is given
to those with medium number of iterations. I'll try a linear ratio
next. Currently we start properly lagging at ~2^20 iterations, which
is why I'm looking for a better way to draw and compute escapes.
