```bash
uvx gha-tools autoupdate ./.github/workflows/ --pin=all --version-strategy=specific --write
```

Calling `lru_cache` on instance methods causes a leak. Fixed as suggested in https://rednafi.com/python/lru-cache-on-methods/

* Initial plan

* Fix issue #1554: Preserve custom thumbnails in gallery script

Modified scripts/generate_gallery.py to check if a thumbnail file already
exists before extracting from notebooks. This prevents the script from
overwriting manually-placed custom thumbnails in the _thumbnails directory.

The gen_previews() method now:
1. Checks if self.png_path exists
2. If yes, logs info message and returns early (preserving custom thumbnail)
3. If no, proceeds with extraction from notebook or uses default logo

Fixes #1554

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

* Use lapack instead of `scipy_linalg.cho_solve`

* Use lapack instead of `scipy_linalg.lu_factor`

* Use lapack instead of `scipy_linalg.solve_triangular`

* Add empty test for lu_factor

* Tidy imports

* remove ndim check

Pre-commit will not install hooks by default, and Copilot has no network access to install afterwards. Pre-install during setup

Co-authored-by: ricardoV94 <28983449+ricardoV94@users.noreply.github.com>

Co-authored-by: Carlos Trujillo <59846724+cetagostini@users.noreply.github.com>
Co-authored-by: Ricardo Vieira <28983449+ricardoV94@users.noreply.github.com>
Co-authored-by: Jesse Grabowski <48652735+jessegrabowski@users.noreply.github.com>

Co-authored-by: copilot-swe-agent[bot] <198982749+Copilot@users.noreply.github.com>
Co-authored-by: ricardoV94 <28983449+ricardoV94@users.noreply.github.com>

Co-authored-by: copilot-swe-agent[bot] <198982749+Copilot@users.noreply.github.com>
Co-authored-by: ricardoV94 <28983449+ricardoV94@users.noreply.github.com>

When a subgraph with multiple outputs is "implicitly" claimed, it can change the dependencies of remaining nodes. A node that depended only on a subset of the subgraph outputs now depends on all of them. Not taking this into account could lead to circular dependent Composites

* Initial plan

* Fix pt.flip to handle negative axis correctly using normalize_axis_tuple
Co-authored-by: ricardoV94 <28983449+ricardoV94@users.noreply.github.com>

* Expand existing test_flip function to include negative axis tests
Co-authored-by: ricardoV94 <28983449+ricardoV94@users.noreply.github.com>

* Fix mypy error by using separate variable for normalized axis
Co-authored-by: ricardoV94 <28983449+ricardoV94@users.noreply.github.com>

* Fix ruff formatting issues
Co-authored-by: jessegrabowski <48652735+jessegrabowski@users.noreply.github.com>

---------
Co-authored-by: copilot-swe-agent[bot] <198982749+Copilot@users.noreply.github.com>
Co-authored-by: ricardoV94 <28983449+ricardoV94@users.noreply.github.com>
Co-authored-by: jessegrabowski <48652735+jessegrabowski@users.noreply.github.com>