Allow a general charset "switcher"

[jferard]

Imported from CSVSniffer. Created at: "2016-12-21T20:32:51Z"

Currently, InputStreamUTF8OrByteCharsetReader falls from UTF-8 to a one byte/one char encoding if it finds a non UTF-8 char. But the implementation (a finite state machine) allows to handle more complicated cases, knowing that a one byte/one char encoding will never fail. A general "switcher" could be easily coded.

Thanks to OH for the idea.