Dataset versioning

[tdudgeon]

We may want to implement versioning of datasets, but this needs thought through first.
It is not currently a priority.

• Datasets are immutable so some mechanism to update them will be useful. A simple numerical versioning system will allow this.
• An operation on a dataset will be an operation on a particular version of that dataset.
• In a shared project different versions of a dataset can be created by different users. Hence a dataset doesn't really have a single owner.
• When you get info on a dataset the versions that are present will be listed.
• When you fetch a dataset you get the latest version unless you specify a specific version.