Revision 627702

Fix problem with alternate Unicode code point overriding existing better ones in CodePointMapping (ex. a char code for NBSP was used in place of SPACE for non-standard encodings).
Made PFM completely optional if an AFM is available. Widths and Kerning are now also read from the AFM. Fallbacks for missing values are in place. If both AFM and PFM are available, both are used to get the best possible result for certain metrics.