[BUG] bjdata (UBJSON backend) prints a NumPy 1.x/2.x ABI notice under NumPy 2.x

Summary

Since the py-ubjson → bjdata migration (#1849), autobahn's optional
"ubjson" serializer is backed by bjdata. bjdata (0.6.6) ships sdist only
and builds its _bjdata C extension against NumPy 1.x via the deprecated
oldest-supported-numpy. Imported under NumPy 2.x (now the default), NumPy
prints a multi-line diagnostic to stderr:

A module that was compiled using NumPy 1.x cannot be run in NumPy 2.4.x ...
  from _bjdata import dump, dumpb, load, loadb
AttributeError: _ARRAY_API not found

import bjdata still succeeds and UBJSON (de)serialization of normal WAMP
payloads works — only bjdata's unused NumPy-NDArray path is affected — but the
notice is alarming and shows up in any app that imports
autobahn.wamp.serializer with bjdata installed (e.g. crossbar version).

Root cause / where the real fix lives

This is a bjdata packaging bug (it should build against numpy>=2.0, which
is dual-compatible, and ship wheels). Filed upstream: https://github.com/NeuroJSON/pybj/issues/7
NeuroJSON/pybj (see also the existing NeuroJSON/pybj#6, which is a different
topic).

What autobahn could do (options, none urgent)

1. Document the caveat for the serialization/all extras (UBJSON backend
   is CPython-only and currently emits a NumPy-2 notice until bjdata is fixed).
2. Pin bjdata to the first release that builds against NumPy 2.x, once
   available.
3. Nothing in code — autobahn cannot fix bjdata's build; this is mainly a
   tracking note so the ecosystem is aware.

Downstream note

crossbar has added a short-term, narrowly-scoped stderr filter for this
specific benign notice in its CLI entry point, to be removed once bjdata ships a
NumPy 2.x build. (crossbar issue: see the crossbar tracker.)

Related: #1849.

Checklist

• I have searched existing issues to avoid duplicates
• I have provided a minimal reproducible example
• I have included version information
• I have included error messages/logs

Start with the optional bjdata import in autobahn.wamp.serializer, then review NeuroJSON/pybj#7 and the related crossbar tracker entry. The issue presents documentation, dependency pinning, or no-code tracking as options, so completion requires a maintainer decision and depends on an upstream NumPy-compatible bjdata release.