Change ItemDictionary<T>.GetEnumerator() to iterator

[Microsoft.Build.Collections.ItemDictionary\.GetEnumerator()](https://github.com/dotnet/msbuild/blob/96a618ca59610908eddc92bfd0198746c34f0ed5/src/Build/Collections/ItemDictionary.cs#L154-L160) currently creates an instance of a nested Enumerator class. It could be rewritten as an iterator method with nested `foreach` loops and `yield return`.

Advantages:

- Source code becomes shorter and easier to understand.
- Ensures that Dispose is called at the appropriate places. (But it looks like a no-op anyway.)
- Gets rid of the custom IEnumerator.Reset() implementation that has a bug ().

Risks:

- Iterators throw NotSupportedException if IEnumerator.Reset() is called, unlike the current custom enumerator. However, the current implementation of IEnumerator.Reset() has a bug already, and ItemDictionary\ is internal, so perhaps nothing calls Reset.
- The [current implementation of IEnumerator.Current](https://github.com/dotnet/msbuild/blob/96a618ca59610908eddc92bfd0198746c34f0ed5/src/Build/Collections/ItemDictionary.cs#L422-L427) throws InvalidOperationException (by attempting to read `((IEnumerator)_listEnumerator).Current`) if there is no current element. An iterator would silently return the previous value of IEnumerator.Current instead. If some caller incorrectly reads IEnumerator.Current in such a situation, then an iterator might make that bug more difficult to find.
- Runtime performance might become slower, or faster.
- Debugging might become more difficult, as the compiler-generated state machine would be more complex than the source code.