fix: auto-detect Jupyter kernel to avoid blank trame renders

[dipinknair]

Bug: plot() renders blank after installing pyvista[jupyter]

Root cause

Installing pyvista[jupyter] sets PyVista's global Jupyter backend to "trame".
PyVistaInterface.show() passed jupyter_backend=None straight through to
pv.Plotter.show(), so it silently inherited "trame", which tries to spin up
a local trame server. VS Code notebooks cannot render the resulting widget —
producing a blank output and a JS SyntaxError: Unexpected token ':'.

Fix

In PyVistaInterface.show() (pyvista_interface.py),
before calling pv.Plotter.show(), detect whether the code is running inside a
Jupyter kernel and default jupyter_backend to "static":

elif jupyter_backend is None:
    try:
        from IPython import get_ipython
        _ip = get_ipython()
        if _ip is not None and "IPKernelApp" in _ip.config:
            jupyter_backend = "static"
    except Exception:
        pass

"static" renders an inline PNG — no server required, works in VS Code,
JupyterLab, and Spyder.

Behaviour after fix

Scenario	Backend used
Doc build (USE_HTML_BACKEND)	"html" (unchanged)
Jupyter kernel, no explicit backend	"static" (new default)
User passes jupyter_backend="trame"	"trame" (explicit override respected)
Plain script / Qt	unchanged pass-through

[RobPasMue]

With this change we would force static renderings on Jupyter NB. And PyVista (via trame) supports interactive ones. Let's try to figure out why it's not rendering on the user's machine.

We could have some kind of fallback though in the sense that if the render goes blank (i.e. fails) we attempt to render it statically. But not by default I'd say. Thanks for investigating @dipinknair

@AlejandroFernandezLuces can you sync with @dipinknair to investigate the issue? He'll share details on the user's failure

[AlejandroFernandezLuces]

Sure, I'll check on my end

[AlejandroFernandezLuces]

Just for the record, this is an issue we have had before with PyVista. Since the change to Trame, visualization in Jupyter and in our docs by extension has been difficult. Using PyVista and following the guide they have for using Trame and Jupyter, I'm not able to render anything on screen, I just get the loading screen:

The alternative that we can do is to fall back to HTML backend when in a Jupyter environment. It's does not have the full Trame functionality but at least it has a basic interactive 3D scene.