Corrade::TestSuite::Comparator<long double> class

Fuzzy-compare for long double values.

Uses comparison algorithm from http://floating-point-gui.de/errors/comparison/ with epsilon equal to 1.0e-17l on platforms with 80-bit long double and 1.0e-14 on platforms where it is 64-bit. This is always one digit less than how Utility::Debug or Utility::format() prints them. Unlike the standard floating-point comparison, comparing two NaN values gives a true result.