Warn for fields declared with generic management or with generic domain type

[mppf]

This PR adds a warning for two cases:

• fields declared like var x: domain. We want users to write var x: domain(?) to emphasize that domain is a generic type and therefore x is a generic field.
• fields declared like var x: C, where C is a class name. These have generic management and cause the field to be generic but that is often unexpected and leads to hard-to-understand errors. We want users to specify memory management explicitly to avoid this kind of genericity. This PR does not introduce a warning for the case where memory management is specified, however the class is generic (that is future work).

It adds these warnings in checkSurprisingGenericDecls which I plan to extend in the future for warnings for variable declarations like var x: someGenericRecord;.

For https://github.com/Cray/chapel-private/issues/5031

Some of this PR is based upon earlier work in PR #21404.

Reviewed by @vasslitvinov - thanks!

• full comm=none testing

[bradcray]

Unless I missed a PR, doesn't this:

fields declared like var x: domain. We want users to write var x: domain(?) to emphasize that domain is a generic type and therefore x is a generic field.

make domain fields inconsistent with other generic fields, as in this example:

record R { 
  type t;
  var x: t;
}

record S {
  var r: R;  // no warning about R being generic, needing to be spelled `R(?)`
  var s: range;  // maybe OK since range is fully-defaulted?
}

var myS = new S(new R(int), 1..10);
writeln(myS);

Put another way, I thought we would want to start requiring domain(?) in all contexts where we might start requiring (?) to indicate a known-to-be-incomplete type. But with this PR, I think it's become a singleton case where we're requiring it—is that right? (that said, if/when other in-progress PRs are merged, it would cease to be special in this regard?)

[mppf]

@bradcray - yes, that's accurate in terms of var r: R vs var x: domain being inconsistent. I am expecting that we will end up with this warning at least for fields (in both cases). It is easy enough to adjust if we decide something different about domain(?), though.