Various fixes and enhancements (incl c3po compatible output)

[willp-bl]

• Fix IOException on files >64kb
• Add option to save serialised metadata objects to a sequence file in HDFS (one per input arc)
• Add option to save characterisation metadata (in a text file) to a zip file in HDFS (one per input arc) - these files should be c3po/tika input compatible although this is not yet tested (testers welcome!)
• Add option to save files that caused the tika parser to crash in the above zip file
• Various fixes and tidying up