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>

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

PyTensor Variables cannot be called `bool` upon

* Add mlx extra ops

* Add log_softmax dispatch for mlx

* Fix split bug, add tests

* Feedback

* Initial plan

* Rename ScalarMaximum/ScalarMinimum to Maximum/Minimum

* Apply ruff formatting fixes

* Remove custom names when class name matches desired name

* Remove scalar_ prefix from log1mexp, xlogx, xlogy0 and fix numba imports

* Fix xtensor test to use backward compat aliases in skip list

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

* Rename `mul` -> `multiply`

* Rename `sub` -> `subtract`

* Space... the final frontier

* `linalg.eig` always returns complex dtype

* Update Eig dispatch for Numba, Jax, and Pytorch backends

* Clean up `pytensor.linalg.expm` and related tests

* Add JAX dispatch for expm

* Implement L_op instead of grad in `Eigh`

---------
Co-authored-by: Jesse Grabowski <jesse.grabowski@readyx.com>

This isn't strictly needed but it's a more intuitive placement

Helpers before dispatchers

They are actually defined in tensor/math.py, but this is better than being in `basic.py`