Evolution fails with 'special' grids

[cschwan]

There's bug in the sliced evolution when the EKO contains Q2 slices which are very close to each other, but still 128 ULPS apart. This is similar to #223.

[cschwan]

Patches have been added to new v0.7.x and v0.8.x branches and also to master in commit 38fc97d.

[cschwan]

We need a test that checks for a regression.