Changed Newton’s method to algebraic solution

[DmitryBaranovskiy]

I replaced the code with algebraic solution. Not only in increases precision, but apparently it is faster.

Old benchmark on my Macbook Pro looks like this

BezierEasing: instanciation x 18,211,751 ops/sec ±0.36% (93 runs sampled)
BezierEasing: call x 34,590,461 ops/sec ±0.18% (102 runs sampled)
BezierEasing: instanciation + call x 7,468,345 ops/sec ±0.26% (94 runs sampled)

The new one looks like this:

BezierEasing: instanciation x 40,143,156 ops/sec ±2.39% (96 runs sampled)
BezierEasing: call x 98,549,765 ops/sec ±0.22% (97 runs sampled)
BezierEasing: instanciation + call x 11,811,166 ops/sec ±1.21% (96 runs sampled)

So, I consider it a serious improvement. You can read about method in my blog post: http://dmitry.baranovskiy.com/bezier-easing.html

Thank you for writing a great library and inspiring my work. I hope, you will accept this change even though it’s quite dramatic.