Simplify away duplicated spelling of flat initial layer depths.

[axch]

As noted in https://github.com/edoddridge/aronnax/pull/110#issuecomment-295968257, both the Fortran core and the Python wrapper implement the ability to initialize a uniform-depths simulation; and the test cases redundantly invoke both mechanisms.

Things we might do:

• Remove that defaulting capability from the Fortran core, in order to simplify it and all the parameterization files.
  • Or choose not to, on the grounds of wanting to ship a plausibly flexible core for the benefit of a hypothetical user that might want to run it in isolation.
• Remove the redundancy from the test case specifications, in favor of using either initHfile or hmean consistently in all circumstances that can be spelled with either.

[edoddridge]

The bathymetry input supports a similar shortcut in the Fortran core as well.

I'm not sure that we should remove the functionality from the core. For the hypothetical user who doesn't want the python wrapping, it does provide a useful shortcut. Actually, what I mean is that I'm not sure we should prioritise above other tasks.

However, in both cases I think it would be better to move the variables to the same section of the config file as the filename string. That will at least hint strongly that the two are linked. We could also rename the variables to something more descriptive than hmean and H0