rename_dims to an existing dimension name

```
ds = xr.Dataset({"a": ("x", np.arange(10)), "b": ("y", np.arange(10))})
ds.rename_dims({"x": "y"})
```
```
ValueError: Cannot rename x to y because y already exists. Try using swap_dims instead.
```

This should be possible since both `x` and `y` have the same lengths. 
Thoughts?

For indexed dimensions, should we check that the indexes are equal?
