Unify FM over all grids

[tcclevenger]

Store a single FieldManager for all fields over all grids.

This solves the issue in #6789 where having a group as a subset of another, and allocated on different grids, was causing subview indices mismatch between grids.

Also

• Remove Derivation type. Now groups contain the union of all fields registered over all grids. I.e., requesting "tracers" on GLL grid without registering any tracers on GLL will give all tracers registered on PG2 grid.
• Remove "Preferred" bundling. You either require bundling or not.

[tcclevenger]

A few comments highlight the important pieces since this is such a large PR. So many of the changes come from every instance of the FM has to be used differently (passing grid names instead of only one grid being stored).

[tcclevenger]

Quite a few fails in the CI, odd some didn't trigger on my workstation (rrtmgp standalone for instance)... Leaving Draft for now.

[bartgol]

I have some preliminary thoughts.

[bartgol]

Looks good, but I think there are a few things to fix.

[bartgol]

This method could have kept the original signature, since in this case we do return all of the groups. My comment was only for the FM method, which returns the info only for one group. Not important though.

[tcclevenger]

Ahh sorry, just did a massive replace with sed.. Forgot this method existed, I'll change it back.

[tcclevenger]

@bartgol Had to correct the equivalent_layout() function in the tensor case, was caught in MPASSI case. Should be good to go.

[bartgol]

@mahf708 brought this up in a private conversation. I think this is the kind of places where std::optional may be beneficial. RIght now, get_tensor_dims errors out if the layout is not that of a tensor. So before trying to grab the tens dims, we must first check if the layout is indeed that of a tensor. Instead, we could think of just returning std::optional<std::vector<Int>>, with the optional NOT storing a value for non-tensor layouts. Then, the calling code would have to handle the potential error:

auto tensor_dims = template_layout.get_tensor_dims();

Any code that access tensor_dims for non-tensor layouts (without doing some preliminary checks) would get a std::bad_optional_access exception thrown.

I'm not advocating for std::optional, but since Naser brought it up just yesterday in a private convo, I thought I mention it, to see what your opinion is. I think it could potentially make the calling code shorter, removing the responsibility for checks from the provider to the customer (in this case, from the FieldLayout methods, to AbstractGrid::equivalent_layouts). Since the calling function may already make sure that the return value should be fine (in this case, via the switch), it could be cleaner:

auto tensor_dims = template_layout.get_tensor_dims();
...
switch (template_layout.type()) {
...
  case LayoutType::Tensor2D:
        ret_layout = get_3d_tensor_layout(midpoints, *tensor_dims, tdims_names);

I haven't made up my mind yet, but maybe worth thinking about it.

[bartgol]

Note: we could also just access the tens dims inside the switch cases that correspond to tensors. I think the reason I did not do that, is that whenever you decl variables local to the case, you need to add a scope, and it looks "odd" when you have a scope for one case and not for others (and it's a lot of extra lines to add the scope for all cases):

switch (a) {
  case A1:
     ...
  case A2:
  {
     auto local_var = ...;
  }
  case A3:
     ...
}

[tcclevenger]

Yeah, I don't like the scope in the cases, and then for the vector and tensor quantities you need to repeat code (but the way it's done here we need the extra ? : line).

I don't have experience with std::optional but it sounds like it's meant for this situation. It does make the func signatures more complicated whereas returning an error if we aren't in a tensor case is probably easier to understand to a newer C++ coder. But I need to look more at std::optional to really understand what the changes would be.

I'll merge as is, but may consider it with the follow up I plan to do.