Simulation parameters and prettier show methods

[ali-ramadhan]

This PR allows simulations to hold parameters, useful if you need to access stuff from the progress function (e.g. see ali-ramadhan/Atmosfoolery.jl#63) going with the idea that we'll probably get rid of model.parameters in favor of more local parameters.

It also cleans up show functions for models and simulations.

[codecov]

Codecov Report

Merging #639 into master will not change coverage by %.
The diff coverage is n/a.

@@           Coverage Diff           @@
##           master     #639   +/-   ##
=======================================
  Coverage   78.11%   78.11%           
=======================================
  Files         119      119           
  Lines        2326     2326           
=======================================
  Hits         1817     1817           
  Misses        509      509           

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 7b5cc97...ff5ff0b. Read the comment docs.

[glwagner]

Yes. Let's get rid of model.parameters ASAP. BoundaryFunction also needs parameters.

[glwagner]

BTW, the progress function does not have to run on the GPU, which means that flexible callable objects can be used which have embedded parameters. This is the way I envision progress function development moving forward. We may want eventually to have complex progress functions for fancy logging / in-terminal plotting, eg. A simulation.parameters field is probably not the right approach --- we want callable objects, I think.

[ali-ramadhan]

Ah interesting yeah a callable object would have worked as well. We should change
https://github.com/climate-machine/Oceananigans.jl/blob/7b5cc97c18c8cd7dcc7a866312630df511735189/src/Simulations.jl#L158
to

iscallable(f) = !isempty(methods(f))
...
iscallable(sim.progress) && sim.progress(sim) 

See: https://discourse.julialang.org/t/how-to-determine-if-an-object-is-callable/2897

Although to have a callable object with parameters you'd have to define an new type right? (As in https://docs.julialang.org/en/v1/manual/methods/#Function-like-objects-1)

Either way, would be good to move the discussion to an issue on getting rid of model.parameters.

[glwagner]

Would a blacklisting strategy work better than whitelisting? What kind of progress object are we defending against? (Is it just nothing?)