Bug: needextract does not support "layout" anymore

[danwos]

Setting layout in needextract does not have any effect. Instead, the original layout is taken.

You can see this also in our documentation:
https://sphinx-needs.readthedocs.io/en/latest/directives/needextract.html#layout

The results do not use the set layout focus_r from above.

[chrisjsewell]

@danwos to note, I'm not sure if this ever actually worked 😅; there is an old version 1.2.0 build still available on RTD, and you can also see it does not work there: https://sphinx-needs.readthedocs.io/en/rtd_jaav/directives/needextract.html#layout
its certainly not a new problem (more than a year old)

I've found why, but don't have a solution available yet.
Here:

sphinx-needs/sphinx_needs/layout.py

Line 106 in c394b1f

env.resolve_references(node_container, docname, env.app.builder) # type: ignore[arg-type]

The need_container looks like this initially: <container><Need>...</></>,
but then because this now gets processed by all the post-transforms, including the one to transform needs, afterwards it looks like:
<container><container classes="need_container" ids="SNCB-c3f165af"><table ...,
i.e. you have already turned it in to a table,
so then when you run build_need in the next lines, it doesn't do anything because the need is already built.

In #794, you actually added a commented out line: # node_container += needs[need_id]["need_node"].children.
This feels more like what is meant to happen, i.e. only the children should be "re-processed", not the whole need node.