[ci skip]

This is required for conversion to NEURON. The members themselves do not
need to be included.

The schema does not require the `name` attribute in `Segment`. However,
the NEURON code generator seems to require it:

```
Exception: java.lang.NullPointerException: Cannot invoke "String.equals(Object)" because the return value of "org.neuroml.model.Segment.getName()" is null
```

… `Exception`

Specific errors are better than generic ones.

This is required to ensure that segment groups have already been defined
before they are included in the default ones.

No need for it to be limited to the utility function.

instead of lxml one.

loaders <-> utils, and partial imports cause errors.

prevent an API break

…rivate

Users are not expected to use these.

So that each component has the necessary members. Users are not expected
to know that these are inherited from `GeneratedsSuperSuper` and go look
there.

…otected

Not private---they need to be inherited by all the component type classes

Before, all segment groups were being created as unbranched with the
neuro lex ID being set for a section. However, it is not compulsory that
users will only create unbranched segment groups. For example, a segment
group for "apical dendrites" would not consist only of unbranched
segments.

BREAKING CHANGE: argument of the cell helper functions has changed from
`group` to `group_id` to make it clear to users that it is the id of the
group that is required there, not the object.

The cell builder utilities have not yet been released, so while this is
a breaking change, this will not affect users unless they've been using
the development branch.

So that all component type classes have the common `add`, `info` etc.
methods listed directly.

Feat/doc tweaks

TODO: Really have to figure out why this is always wrongly generated

`get_members` -> `_get_members`

It is sufficient to only include the segment groups in the default
segment groups. This results in shorter NML files.

…ring

When adding lots of segments, one may want to wait until the end of
the process to reorder the groups, instead of reordering after the
addition of each segment (which is an unnecessary use of resources).

We were inferring the segment type from its group id, but this is not
appropriate in cases where a user may not be passing a group id at all.

So, we uncouple the segment type from the group id and ask users to
include it explicitly.

It will run the checks again, which we don't need.

Checking for membership is not performant for segment groups with larger
numbers of segments. Instead, we will add a function to optimise segment
groups that will de-duplicate them etc.

…s are present

There's no need to use `add` here since we're constructing the various
object and can be quite sure that they are correct.

Not using `add` here gives a huge performance boost because `add` gets
members each time, and then goes through them to see where the component
should fit.

very very useful when repeated calls are being made to methods that use
`_get_members` like `_add` and so on.

We can't import `typing` in nml.py at the top level, so adding hints
that require the `typing` module causes trouble. Just remove them.

This particular docstring is always formatted wrongly by black. Not sure
what's going on.

Feat/add sectionise method

- only create all and soma default groups
- dendrite/axon groups created whenever a segment of their respective
  type is added for the first time.

…ions

- adds options to allow disabling when required

… elements are set