IMPROVEMENTS

Flume Source

1. Process event's batch instead of single event.
2. Custom events' headers.
3. Actually we keep a map of processed files. If agent stops while processing file, file is not saved into map, so restarting flume will process file from line 1. Not a bad behaviour, but and interesting improvement.
4. Make configurable recursively process files under maing working directory. Actually is true by default(more info).(DONE).
5. Set a criteria for validating events.Actually we are validating StateEvent under VFS2. Try to set a custom criteria configurable via properties, for example: csv file?, minimun number of columns...