Adopt adaptive-triangulation 0.3.1: batched tell_pending, Rust defaul…

…t_loss, CI job (#499)

adaptive-triangulation 0.3.1 fixes the degenerate-simplex divergence noted
when the backend was introduced (curvature losses crashed on collinear
input) and 0.3.0 added two APIs that move LearnerND's remaining Python hot
loops into Rust. This bumps _MIN_RUST_VERSION to (0, 3, 1) and adopts both:

- tell_pending asks the triangulation for all simplices containing the
  pending point in one simplices_containing call (with the known containing
  simplex as a hint) instead of looping point_in_simplex over every
  neighbouring simplex; the pure-Python backend keeps the original loop.
- LearnerND's default loss_per_simplex is the Rust default_loss when the
  Rust backend is active (same embedded-simplex-volume computation).

Together: 1.10x (2D ring_of_fire, 3000 pts) to 1.42x (3D, 1500 pts) faster
end-to-end on top of the existing Rust backend, sampling identical points
(asserted by a new test).

Also adds the Rust-backend CI job deferred when the backend landed: a nox
session running the suite with ADAPTIVE_TRIANGULATION_BACKEND=rust, now
that the full suite passes with it (294 passed; previously 7 known
curvature-loss failures).

Default to loky executor on all platforms (#497)

Python 3.14 changed the default multiprocessing start method on Linux
from "fork" to "forkserver" (python/cpython#84559), so workers of the
stdlib ProcessPoolExecutor now re-import __main__ instead of inheriting
it. Functions defined interactively (notebooks, executed doc pages) are
pickled by reference and fail to unpickle in the workers with e.g.:

    AttributeError: module '__main__' has no attribute 'sphere'

loky's reusable executor serializes functions by value via cloudpickle
and was already the default on macOS and Windows (and loky is already a
required dependency), so use it on Linux too. This also removes the
now-dead lambda-pickling check in AsyncRunner that only applied when the
default executor was ProcessPoolExecutor.

Raise myst-nb execution timeout to fix flaky Read the Docs builds (#495)

The logo animation notebook (docs/source/logo.md) renders 360 frames and
encodes them with libvpx-vp9, which takes close to 3 minutes on Read the
Docs builders. The 180 s nb_execution_timeout was intermittently exceeded,
causing docs builds to fail with CellTimeoutError.

Fix AsyncRunner failure on Python 3.14 due to asyncio.get_event_loop(…

…) removal (#490)

Fix several deprecation warnings (#481)

Update changelog to v1.4.0 (#480)

Fix readthedocs.yml (#474)

Ref: https://about.readthedocs.com/blog/2024/12/deprecate-config-files-without-sphinx-or-mkdocs-config/

fix scipy deprecation warning (#465)

Release v1.3.0 (#459)

Update CHANGELOG.md for v1.2.0 (#454)

* Update CHANGELOG.md

* Update CHANGELOG.md

* Update CHANGELOG.md

---------

Co-authored-by: github-actions[bot] <github-actions[bot]@users.noreply.github.com>