Deserializer with buffer size

[ivanprado]

Currently we are using the Hadoop Deserializer class for the custom Pangool serialization: http://pangool.net/userguide/custom_serialization.html

But some serialization mechanisms needs the buffer size in order to deserialize properly (i.e. ProtocolBuffers). With current API the serialization needs to write the buffer size to the disk... But that is something we have already done with Pangool!. So we would be writing the buffer size twice.

That can be avoided by passing the buffer size to the deserialize() method. That would mean that we would have to stop using the Hadoop Deserialize class and use our own new interface for that.