Clarify management of coordinate systems

[PProfizi]

The result.coordinate_system operator currently returns a field with 3x3 rotation matrix and origin coordinates for a given coordinate system ID.
This is now being exposed in other places such as the cyclic support (see #153).

This means the CS location should at least be made available and documented and the management of coordinate systems should be clarified.

Do we need a CoordinateSystem dedicated object? Do we need to manage the CS location in some other places? Do we need to allow to add coordinate systems to plots?

It could be argued that the location of the field should be overall and that the fact it represents a coordinate system could be communicated another way:
If for example we want to define the field of element coordinate systems, I would expect an Elemental field with the aforementioned matrix at each entity. Same for a field of coordinate systems at each node, at each part... etc.

Note that all operators in geo related to coordinate system conversion seem to use a unique format.

Note: latest PRs server-side now use a GenericDataContainer for coordinate system outputs. See the new coordinate_system_data_provider.

We will need to properly deprecate the current behavior of the result.coordinate_system so it returns a GDC we could then wrap in a dedicated Python object.

@oparreno @rafacanton any input on this?