Redesign the file storage to use MappedByteBuffer

[bdezonia]

The file based storage for types would be much improved if it can be based upon nio's MappedByteBuffer. Whatever is implemented needs to be able to handle huge files and to do so in an efficient manner. Maybe all data types will change from working with RandomAccessFile to something else and thus this might be an api breaking change.

[bdezonia]

For the 1.0.0 release I decided to settle on ByteBuffer enabled code (and not yet MappedByteBuffer). The code performance is now 10 to 100 times faster than before. If further speed is necessary I will revisit the MappedByteBuffer idea.