[cairo] Analytical curve bounds

[Behdad Esfahbod]

Kalle Vahlman wrote:
> The fills seem legit though, since there's plenty of them in the
> positive side and the new code does get called in them in abundance.
> For example, during fill_solid_rgb_over-64 I counted (well, wc did ;)
> 405 calls to _cairo_spline_bound().
> 
> So looks like the new code should be a definitive win.

Great.  Would be even better if you test with master as I pushed some
optimization in the code to avoid calling sqrt in most cases.  That may be
loss with modern processors, donno.

behdad