* Force environment dependent xfail test to fail

* New version of MLX crashing in CI

This way they can be fully cached when re-running tests

Direct import is not properly mocked by tests when trying to run `compare_numba_and_py` with `eval_obj_mode=True`

These tests were covering things that don't exist anymore. params in python perform method of Ops, or misbehavior of an Op not respecting the signature

This is failing in the CI and isn't used?

Using DeprecationWarning to keep it visible only for devs for now

return_steps has not been a thing for 14 years

Partially reverts d8943500

Bug introduced by 671a821d

Bumps [actions/download-artifact](https://github.com/actions/download-artifact) from 5.0.0 to 6.0.0.
- [Release notes](https://github.com/actions/download-artifact/releases)
- [Commits](https://github.com/actions/download-artifact/compare/634f93cb2916e3fdff6788551b99b062d0335ce0...018cc2cf5baa6db3ef3c5f8a56943fffe632ef53)

---
updated-dependencies:
- dependency-name: actions/download-artifact
  dependency-version: 6.0.0
  dependency-type: direct:production
  update-type: version-update:semver-major
...
Signed-off-by: dependabot[bot] <support@github.com>
Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com>

* Fix Blockwise vmap dispatch for no batch dimensions

Updates comments in funcify_Blockwise to avoid confusion about behaviour. Adds tests to verify correct behavior for these cases.

* pre-commit

* Take our docs

Co-authored-by: jessegrabowski <jessegrabowski@gmail.com>

Also remove legacy `__div__`

We should instead install the GitHub App. (I don't have permissions to do this.)

See the announcement:
<https://about.readthedocs.com/blog/2025/06/announcing-our-github-app-beta/>

* Initial plan

* Remove InRange scalar op and add missing __rtruediv__ and __rfloordiv__

* Remove __rtruediv__ and __rfloordiv__ changes per review feedback

---------
Co-authored-by: copilot-swe-agent[bot] <198982749+Copilot@users.noreply.github.com>

Using JAX Scan machinery to create MIT-SOT, SIT-SOT, and NIT-SOT buffers for us seems to be more performant than working directly on the pre-allocated buffers and reading/writing at every iteration. There is no machinery to work with MIT-MOT directly (just like in PyTensor user-facing Scan).

Co-authored-by: Jesse Grabowski <48652735+jessegrabowski@users.noreply.github.com>