Allow changing nseg after initialization

[michaeldeistler]

This PR is a follow-up to #418 and #426. It allows to set nseg after initialization of the module. This is achieved with a new set_ncomp method. This PR is the final requirement for a new release with these features.

API

Works for branches:

comp = jx.Compartment()
branch = jx.Branch(comp, nseg=2)
branch.set_ncomp(4)

...for branches within a cell:

comp = jx.Compartment()
branch = jx.Branch(comp, nseg=2)
cell = jx.Cell(branch, parents=[-1, 0, 0])
cell.branch(1).set_ncomp(4)

...and for branches read from SWC files, in which case it reuses the xyzr to compute optional radiuses:

cell = jx.read_swc(fname, nseg=2)
cell.branch(1).set_ncomp(4)

Assumption and Raises

The set_ncomp function will raise an error if:

• There are stimuli in any compartment in the module
• There are recordings in any compartment in the module
• The channels of the compartments are not the same within the branch that is modified
• The lengths of the compartments are not the same within the branch that is modified

In addition, unlike the data was read from an SWC file, .set_ncomp() raises an error if:

• The radiuses of the compartments are not the same within the branch that is modified

Finally, we currently raise NotImplementedError if the Module is a Network, e.g. for

net.cell(0).branch(1).set_ncomp(4)

This is because I do not want to deal with synapse indexing for now.

We also raise a NotImplementedError if the Module is an entire Cell, e.g. for

cell.set_ncomp(4)

[jnsbck]

Cool! Awesome that this is working now on both the API and solver levels now! Amazing effort! I have not specifically thought about how I would implement this, but I assume you have spent quite some time thinking about how to best do it, so I assume its a good solution. I left a few comments. Good to go once comments are addressed.