- `get_code` will keep the previous behavior of `show_code`
- `get_code_str` will be equivalent to previous `str(show_code(...))`

This adds a function for lazy people who want to write

```python
dh = create_data_handling((20, 30))
dh.add_arrays('x, y(9), z')
```

instead of

```python
dh = create_data_handling((20, 30))
dh.add_array('x')
dh.add_array('y', values_per_cell=9)
dh.add_array('z')
```

because in most cases you want to use more than one array.

otherwise the index gets lost

uses new Field.neighbor_vector

move_constants_before_loop needed to be changed to move already-existing variables up if necesesary and to preserve constness when renaming variables

also add a test for subexpressions

It takes a list of assignments to staggered fields and otherwise behaves like create_kernel.
The old create_staggered_kernel was not removed because all the optimizations still need to be ported over.

When storing fluxes on a staggered grid, the usual sign convention is that
fluxes point outward from the cell. Previously, we did not respect that as
staggered_access("E") would return the same thing as staggered_access("W")
would when called from the eastern-next cell. Now, when a field is declared
as STAGGERED_FLUX, it returns an accessor with a prefactor of -1 in that
case. The previous behavior where sign is not reversed is still useful when
e.g. storing sums (e.g. mean values) instead of differenes (e.g. finite
difference fluxes) on the staggered grid.

NW (-1/2, 1/2) and the like were previously mapped to the wrong cell