Parallelising in `z`

[ZedThree]

To support parallelisation in z, we need to do a few things:

• check loops over z that currently go over [0, LocalNz) should be switched to [zstart, zend] instead
• check ranges over z that currently go over [0, LocalNz) should be switched to [zstart, zend] instead
• convert remaining loops over x, y, z to use BOUT_FOR where possible
• guard anything using FFTs on NZPE==1
• switch from using modulo for z periodicity to guard cells
  • add communication in z (even for serial-in-z)
  • default MZG
  • performance tests!
• domain decomposition in z
• modify boutdata and xbout to support parallel z