constraints for to and extensions for addition and getitem (#81159)
- Added constraints for addition that carry *runtime* value. This situation is when addition is between two dimensions. There are more cases to cover that we will add as we go though different HF models.
- Added constraints for "to" which is of the form input = output (propagating the shape)
- Added constraints for getitem when the index is of the form of a tuple. This yields a tensor instead of a number and we would have to propagate the shape instead of the getitem expression rather than it's runtime value.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/81159
Approved by: https://github.com/jamesr66a
