To arguments of `read_grid_inst_eddies`. Because we are going to need
them to check the order of dimensions in SSH variable.

Actually, this does not make much difference since we only produce a
warning and processing continues anyway.

Because we are going to need them to check the order of dimensions in
primary variables.

Preparing to move the calls to `nf95_find_coord` out of the `if` construct.

Preparing to promote them to module variables.

To module `read_grid_inst_eddies`. Makes more sense and is
simpler. `stag_[uv]` and `single_grid_uv` no longer need to appear in
procedure `input_ssh`. `stag_[uv]` no longer needs to appear in main
program unit `test_local_extrema`.

Promote periodic to variable of module
`read_grid_inst_eddies_m`. `periodic` was transferred from
`read_grid_inst_eddies` to `get_var` through `input_ssh` without being
used in `input_ssh`. Note that we now have to call config and
`read_grid_inst_eddies` in program `test_local_extrema`.

Check `max_radius` and nlon in `read_grid_inst_eddies`. Makes more
sense, this is a grid problem.

`Intent(out)` first. We could not do this before because of reference
to periodic.

To module `read_grid_inst_eddies_m`. Because we want to use it in
`get_var` and that would be a circular dependency.

It was computed twice, in `input_ssh` and in `set_all_extr`.

Preparing to promote `copy_max_radius` to module variable, as it is
computed twice, in `input_ssh` and in `set_all_extr`.

Move variable `single_grid_uv` to module
`read_grid_inst_eddies_m`. Makes more sense since `single_grid_uv` is
a property of the grid.

Makes more sense since `stag_[uv]` are properties of the grid.

Makes more sense since periodic is a property of the grid.

Move definition of `single_grid_uv` after call to
`read_grid_inst_eddies`. Preparing to move `stag_[uv]` to
`read_grid_inst_eddies`.

Extract code from procedure `input_ssh` to new procedure
`read_grid_inst_eddies`. Preparing to move procedure `get_var` to its
own module. We could not move `get_var` before because `get_var` uses
nlon and that would be a circular dependency. With the code extracted
from `input_ssh`, we also move to module `read_grid_inst_eddies_m` the
variables `uniform_lon_lat`, nlon, nlat, step, `corner_whole`,
longitude, latitude, `lon_steps` and `lat_steps`. A new namelist is
read: `read_grid_inst_eddies_nml`. Variable `uniform_lon_lat` moves
from namelist `input_ssh_nml` to namelist `read_grid_inst_eddies_nml`.

Instead of two calls to plt functions.

It was already possible for `get_slice` to return None so a user
calling `get_slice` should be ready for this.

We are amoving the NetCDF-file arguments to the end because we are
going to allow a variable number of them.

To a time dimension with arbitrary name.