hdf5 write_copy error

Nobody has claimed this yet.

@AdriaanRol
### Steps to reproduce
1. load a dataset
2. change formatter to HDF5
3. use `dataset.write_copy(path="something")`
### Expected behaviour

it should save tghe dataset in hadf5 format
### Actual behaviour

write_metadata() got an unexpected keyword argument 'read_first'
which gets called in:

```
c:\users\damaz de jong\pycharmprojects\qcodes\qcodes\data\data_set.py in write_copy(self, path, io_manager, location)
684 self.snapshot()
685 self.formatter.write_metadata(self, io_manager, location,
--> 686 read_first=False)
687 finally:
688 for array_id, array in self.arrays.items():
```

My pycharm also tells me that your signature on hdf5_format.py line 234 does not match the signature of the base class `Formatter`
### System

**operating system**
Windows 10

**qcodes branch**
master

**qcodes commit**
35515fa

Start in qcodes/data/data_set.py at write_copy and compare its write_metadata call with the HDF5 formatter signature in hdf5_format.py around line 234 and the base Formatter declaration. Reproduce the Windows HDF5 write_copy steps, then verify that the dataset saves successfully without the unexpected read_first argument error.

Written by the indexing model from the issue text.