added timings: 0: no timings, 1: timings output

spaceparams.F90: in the block-collect subroutines, use 3rd dimension
  of b, not of a

Lots of changes, all in the parallel code.

It was necessary to put many parallel code in the subroutines, 
and I removed the calls to distribution routines between the
subroutine calls.

The most important parallel routines are now xmpi_shift and
xmpi_getrow. The latter is only used in two occasions, but it
is possible that the use of that subroutine is needed in more
places. See xmpi.F90 for descriptions.