Extract axes from SimulationBuilder

[bonjourmauko]

This is a work in progress based on my willingness to give a proper answer to #933 and #934.

I hoped to have a more advanced proposal for a request for comments, but axes are a hard one to tackle properly.

What do you think? Any feedback at this point @benjello @guillett?

Details below.

New features

• Introduce Axis, AxisArray, and AxisExpander.
  • It clearly separates the data logic from the domain logic (setting up a collection of axes coherently vs actually expanding a simulation over a given collection of axes).

Improvements

• Introduce proper separation of concerns between an axis, a multi-dimensional axes collection, and its expansion.
  • Allows for a better readability and testability, and future improvement of the expand axes feature.
  • Adds more unit test coverage, including some untested invariants (like the number of counts for a parallel set of axes).

Deprecations

• Deprecate add_parallel_axis, add_perpendicular_axis, and expand_axes.
  • Those functionalities are now provided by AxisArray and AxisExpander.

[bonjourmauko]

Thanks @benjello, to be more clear what I would like is to know if the assumptions I'm doing are the right ones in terms of field expertise. For example:

• Does it make sense that all axes perpendicular to themselves have the same amount of counts, or steps?
• Does it make sense to have more than 2 dimensions? (a 4-dimension axes collection is unplottable, but maybe it makes sense for a use case of yours).
• Etc.

Most of the intent is not yet implemented but outlined in the documentation.

[benjello]

• Does it make sense that all axes perpendicular to themselves have the same amount of counts, or steps?

• No

• Does it make sense to have more than 2 dimensions? (a 4-dimension axes collection is unplottable, but maybe it makes sense for a use case of yours).

• Yes

[bonjourmauko]

Thanks @benjello, this is not yet ready to be merged, I wanted to be sure I was shooting at the right direction.

• Does it make sense that all axes perpendicular to themselves have the same amount of counts, or steps?

• No

So we could have, for example, something like:

axes[[2,2], [3], [4]]

But no:

axes[[5,2], [3], [4]]

?