Performing a case unsensitive searching for accented characters returns
incorrect matching bounds.

For example, type this in a gedit document:

αινσϊ

Now try searching for α -> the first two characters are shown as matching
Searching for ι show the third and fourth chars as matching.

This is due to UTF8 normalization, which expands accented characters as
compose sequences.

Happens both for forward and backward searches.  Case sensitive searches
work correctly.

And I thought I had fixed searching... :-(