read_csv() restarts index (if not loaded) at every chunk

See [this comment](https://github.com/pydata/pandas/pull/12173#discussion_r51347252). Basically we now have

```
In [8]: read_csv(StringIO(u'A,B\n1,2\n3,4'))
Out[8]: 
   A  B
0  1  2
1  3  4
```

but

```
In [9]: concat(read_csv(StringIO(u'A,B\n1,2\n3,4'), chunksize=1))
Out[9]: 
   A  B
0  1  2
0  3  4
```

We want the latter to also have index (0,1).
