Improvements to dataset_to_dataarray/frame

[OriolAbril]

After trying them out a bit, I think it would be nice to add a few features:

dataset_to_dataarray

• Add a multiindex : bool, default True option. Having the multiindex can be nice as it allows subseting using the original dimensions or variable names, but in some cases it ends up being a nuisance and would be better off with the labels only (which can also be customized)
• Add a new_dim : str, default "label" option. I think "label" is a nice default given we populate its contents with the labeller so I wouldn't make it a required argument like xarray does, but doing something similar to az.summary for example I wanted the new dimension to be metric instead.

dataset_to_dataframe

• Extend multiindex options to allow False (default) no multiindex anywhere, True behaviour unchanged, multiindex everywhere, "row" multiindex for the rows only and "column" multiindex for the columns only.
• (unsure about) allow using a different name for the row multiindex that is not "sample". Unsure because I generally don't use it but would be as happy (or even happier) with that whole column with tuples in it being gone than it still there with a better name