Add scan_indices to Simulations

[james-isbister]

I think it would be useful for us to represent the n-dimensional index of a Simulation in the n-dimensional hypercube of Simulations within a SimulationCampaign. For example:

For each Simulation

scan_parameters= {"param_a": "bar", "param_b": 0.4, param_c: (1, 2, 3)}
scan_indices = {"param_a": 1, "param_b": 3, "param_c": 0}

where the values in the scan_indices dictionary are the indices in the scan_parameters dictionary of the SimulationCampaign:
i.e.

SimulationCampaign.scan_parameters = {"param_a": ["foo", "bar"], "param_b": [0.1, 0.2, 0.3, 0.4, 0.5]}, "param_c": [(1, 2, 3), (4, 7)]}

This is required, for example, for us to generate a bbp-workflow campaign summary file as input to blueetl, on the fly. The proposed scan_indices would allow us to do this during or after staging a campaign (after staging is sufficient for now) irrespective of the layout of the directories that the campaign is staged into.

It also means we don't have infer the locations of individual coordinates within the hypercube, which is:

1. Impossible if the values for a certain dimension aren't unique i.e. "param_a": ["foo", "foo", "bar"]
2. Awkward when the scan values of are of more complex types (i.e. tuples, lists)

In future this might enable other uses, also, such as representing the hypercube of simulations visually

Good to know your thoughts. I'm sure there are other alternatives, but this seems the simplest to me. It would be difficult to do for existing simulation campaigns (especially those where 1 and 2 apply) - so would we leave the values in those columns empty for those rows, for example?