29 save toy cov

Saving the covariance matrix to the hdf file should work.

For the loading, two problems arise:

• as the toy fit result contains several fits, should we still have 1 fit -> 1 fit result? (which makes the use of a simple FitResult.from_hdf method questionable: should this really return several instances?)
• as the fit results are, before being saved as hdf, converted to a normal dict, we cannot recover the parameters simple and loose the information about their ordering, which would be a requirement in order to interpret the covariance matrix (<- this problem could be solved differently by storing the param names in the saved DataFrame). But also the FitResult._result expects an OrderedDict, right?

What do you think, @apuignav