docs: correct false simulator-JAX claims (registration, eager-on-JAX, TransformerNUFFT)

[Jammy2211]

Summary

Corrects the false JAX claims in this workspace's docs — the autogalaxy half of
PyAutoLabs/autolens_workspace#379. All measured against the installed stack.

1. "The simulator handles pytree registration internally" — false

No simulator anywhere registers pytrees: grep -rn register over both autogalaxy
simulators, both autolens simulators and both autoarray bases returns zero
hits. The documented @jax.jit wrap fails immediately with
TypeError ... value is of type Galaxies.

It is also unimplementable as worded — JAX flattens a jitted function's
arguments at trace time, before entering the callee, so a simulator that
registered internally would already be too late. (PointSolver.solve_triangles
in autolens carries the same note at its own call site.)

The correct call is now named: autogalaxy.jax.register_galaxies_classes(galaxies),
added in PyAutoGalaxy#537 precisely because autogalaxy had no public counterpart to
autolens.jax.register_tracer_classes.

2. "eager via_galaxies_from(...) already runs on JAX" — false

Eager with use_jax=True succeeds but returns dataset.data.array of type
numpy.ndarray, not jax.Array.

3. The @jax.jit simulator recipe does not work at all

Even with correct registration it dies inside autoarray
(preprocess.py:153, then array_2d_util.py) on sites that do not thread xp.
Separate library bug, filed separately. The docs now say so and point at the eager
call rather than presenting a recipe that cannot run.

4. TransformerNUFFT — two separate stale claims

• guides/using_jax.py: docs: reduce likelihood_function.py __JAX__ blocks to a pointer to guides/using_jax.py #181 (merged this morning) shipped "TransformerDFT (the
  default). TransformerNUFFT is not JAX-traceable." Inherited from the old
  __JAX__ blocks and backwards.
• interferometer/simulator.py carried an older variant: that TransformerNUFFT
  is pynufft-backed and "the nufftax replacement is a research path".

Per autoarray's own error text, TransformerNUFFT is "the default JAX-native"
transformer (nufftax-backed; nufftax 0.3.1 installed) and
TransformerNUFFTPyNUFFT is "the legacy pynufft backend". Defaults differ by
class: SimulatorInterferometer → TransformerDFT; Interferometer, which a fit
uses → TransformerNUFFT.

Note on why this went unnoticed

Both __JAX Variant__ sections present their recipe inside a ```python
fence — prose, not executed cells — and neither simulator.py is in
smoke_tests.txt. So dataset_jax = simulate(galaxies) looked like a working
call while never running. Same failure mode as the six likelihood_function.py
__JAX__ blocks removed in #181.

Scripts Changed

• scripts/guides/using_jax.py — rewrote __Writing @jax.jit Yourself__ (claims 1-3); fixed the NUFFT sentence in __Custom Likelihood Functions__ (claim 4)
• scripts/imaging/simulator.py — claims 1-3 in __JAX Variant__
• scripts/interferometer/simulator.py — claims 1-3 plus the stale nufftax-as-research-path paragraph

Notebooks, llms-full.txt and workspace_index.json regenerated via
PyAutoHands/autohands/generate.py.

Upstream PR

PyAutoLabs/PyAutoLens#665 — the same stale auto-call claim
in autolens/jax/registration.py's module docstring. Docstring-only with zero API
surface; nothing here imports new API, so merge order does not actually matter.

The register_galaxies_classes function this PR now points readers at is already
merged (PyAutoGalaxy#537, 9594c00b).

Test Plan

• Both edited simulator scripts run end-to-end
• py_compile on every touched file
• Post-edit grep: zero surviving hits for pytree registration internally, already runs on JAX, and both reversed NUFFT claims
• Smoke tests: 12/12 passed

Generated by the PyAutoLabs agent workflow.