Whilst preparing to fix https://bugs.openjdk.java.net/browse/JDK-8196572
it was noticed that now PYCC fails whereas it was just CIEXYZ before.

The upgrade to LCMS 2.9 changed things slightly so that the PYCC case
now showed a maximum error of 214 vs 207 before when using the
open colour profiles.

However that is an enormous allowed error margin since the way the test
works it could be that almost every component of every pixel be off by
almost the entire colour range.

So I examined the resulting images and I've attached :
PYCC.png - the golden image from which the test starts
dst_open.png: the color converted output using LCMS+the open PYCC.pf
dst_closed.png: the color converted output using LCMS+ the closed source PYCC.pf

It should be obvious that the "open" result is wildly different than either the
original or the "closed" result. And the closed result is nearly identical to the original.

So the difference is all in the profile - not the CMS.

I was able to make the test pass again by applying a slight increase in the tolerance,
but the real problem is the profile and the very high tolerance that was already allowed.

The only resolution seems to be to create or obtain a better PYCC.pf
We can/should then reduce the tolerance allowed.