The translation of static fields now uses `VariableAddress` instead of `FieldAddress`. This fixes the logic as well as the "field address without qualifier address" sanity check.
Fixed a problem when the translating the compiler generated constructors that caused some sanity errors (since they have no body, when translating the constructor block fragmentation happened). Fixed this by skipping the translation of the body, if it does not exist (when translating a function).
This commit implements the language-neutral IR type system for C#. It mostly follows the same pattern as C++, modified to fit the C# type system. All object references, pointers, and lvalues are represented as `IRAddress` types. All structs and generic parameters are implemented as `IRBlobType`. Function addresses get a single `IRFunctionAddressType`.
I had to fix a couple places in the original IR type system where I didn't realize I was still depending on language-specific types. As part of this, `CSharpType` and `CppType` now have a `hasUnspecifiedType()` predicate, which is equivalent to `hasType()`, except that it holds only for the unspecified version of the type. This predicate can go away once we remove the IR's references to the underlying `Type` objects.
All C# IR tests pass without modification, but only because this commit continues to print the name of `IRUnknownType` as `null`, and `IRFunctionAddressType` as `glval<null>`. These will be fixed separately in a subsequent commit in this PR.
A constructed type, `C<T>`, where `T` is the type parameter of `C`, is represented
in the database as the corresponding unbound generict type `C<>`. Consequently, the
type conversion library, which only considers `ConstructedType`s, does not handle
all implicit conversions. For example, in
```
interface I<in T1, T2> where T1 : C
```
there should be an implicit conversion from `I<C, T2>` to `I<T1, T2>` (=`I<>`).
