ENH: squeeze - lower dimensionaility of a structure by removing exactly len 1 dims

like numpy.sqeeze but returns the correct pandas object

add in core/generic?

anyone like squash better?

(prob a fair number of special cases where need to check, e.g. empty Series)...
so propose for 0.11

```
def squeeze(self):
    """ squeeze length 1 dimensions """
    return self.ix[tuple([ a[0] if len(a) ==  1 else slice(None)  for a in self.axes ])]
```
