Dumb horrible bug in PAW relativistic calculations

In the new PP-reading code, tag name PP_AEWFC_REL.* was incorrectly referred to as PP_AEWFC_rel.*, leading to aewfc_rel array not being read and to a small but nonnegligible error in the results. Guess what? there isn't a single test with relativistic PAW in the test-suite, that would have allowed to spot the error at once. Some cleanup as well.

Fixes issue #277 (closed). Noticed by Minkyu Park (Univ. of Ulsan, Rep. of Korea)