Feature: hdf5 formatter with json metadata

[peendebak]

This PR adds a new formatters that uses the HDF5 formatter as a base. The metadata is written in plain JSON. In this way the data read back from the metadata is identical to the original data (for all common data types).

• Write metadata as JSON
• Add generic test for all formatters

@AdriaanRol @Unga @core

[peendebak]

@AdriaanRol @jenshnielsen The build system is complaining about the type of the arguments of write_metadata in the HDF5Formatter. Here:

https://github.com/VandersypenQutech/Qcodes/blob/feat/hdf5json/qcodes/data/hdf5_format.py#L239

The io_manager and location have default value None, which is not the case in the parent class Formatter. Do we want to add None as the default value for all the other formatters as well?

[jenshnielsen]

IMHO this is not a big issue as I understand it location and io_manager is required for the json metadata but not for the pure HDF5 formatter so the signatures are "correct" as is. Ways that this could be solved to not raise a style warning

• refactor the common parts of HDF5FormatMetadata and HDF5Format format into a common base class.
• Change the inheritance order and make HDF5Format a subclass of HDF5FormatMetadata pylint/codeacy seems to think it's fine with a subclass with less arguments

but I fail to see how this archives anything other than make the linter happy

[peendebak]

@jenshnielsen I made the linter happy. Long term fix would be to specifiy the arguments of write_metadata and make the formatters obey.