fit inverse parametric directly

We are in control of the residuals, so rather than fit
a parametric form by minimizing (table(x)-parametric(x))^2,
we can fit by minimizing (parametric^-1(table(x)) - x)^2.

This is nice because:
  1) the inverse parametric is really the more important
     curve to fit well, for its use in destinations
  2) it's how we're testing the quality of the fit already

Every diff here is pretty much expected, except perhaps the
Facebook sRGB polyTF changing.  That's from a _slightly_
different d in the parametric fit.  Luckily, it's improved.

Change-Id: If73af229d6175ac866f77dd2b22085c5c3e913aa
Reviewed-on: https://skia-review.googlesource.com/124480
Reviewed-by: Brian Osman <brianosman@google.com>
Commit-Queue: Mike Klein <mtklein@chromium.org>
20 files changed