Revisit decoding of documents from binary to string

[benoit74]

Fix #221

Changes:

• decoding of binary WARC content:
  • do not try all chardet suggested encodings in a row, but only the most probable one
  • if all alternatives to decode have failed, try to simply ignore bad characters when decoding with original encoding indicated in HTTP headers
  • log a warning (with details) when encoding used is not the encoding indicated in HTTP headers or characters had to be ignored

[rgaudin]

LGTM