Consistent approach to storing coordinates

[arsenius7]

According to the spec, cell positions are now stored as x, y, z 1-D attributes; while synapse positions are stored as 2-D attributes (e.g., afferent_center_position).

Ideally we should be consistent here.
Imho, storing positions in 2-D attributes makes sense since we usually need all the coordinates; but we need to recall why we decided to go with x, y, z for cell positions recently. :)

[hernando]

Nobody seems to recall and after having decided to represent orientation with quaternions, which will probably be stored as a 2-D array with 4 columns, it makes sense to reconsider storing positions in a single dataset again.

[arsenius7]

For the moment we decided to get away with 1-D attributes only; and modified the spec for synapse positions accordingly.