BUG: DataFrame.apply axis=1 for str ops with no axis argument

E.g.

```
df = pd.DataFrame({'a': [1, 2], 'b': [3, 4]})
print(df.apply('pct_change', axis=0))
print(df.apply('pct_change', axis=1))
```

both give the output

```
     a         b
0  NaN       NaN
1  1.0  0.333333
```

This should instead raise rather than silently give incorrect results.
