Add reflected tail method for thickness mapping

[cmeyer]

See:

• Allow user to plug-in background and thickness models #17

[bentps]

Should this method replace the current method, or should we keep the current one around? I checked, and the current method gets the ZLP area by integrating the interval around the ZLP peak that is above 1/10 the ZLP value, which fails for instance with Dectris data that has a 1 pixel wide ZLP. My thought is that if the new method is more reliable, it should just replace the current one.