Stephan Rave

This PR can be seen as an alternative approach to #2211. It adds an `orth` interface function that orthonormalizes a `VectorArray` by dispatching to various orthonormalization algorithms. Main characteristics: -...

Vectorize orthonormalization steps in gram_schmidt

3

This refactors the `gram_schmidt` algorithm to orthonormalize all vectors to the right in a single vectorized operation. Depending on the use case, this can lead to significant speedup for `VectorArrays`...

Support Python 3.12

9

This PR adds official support for Python 3.12 and updates ci-current and conda CI environments to this version.

Add MOR methods overview to documentation

7

See https://docs.pymor.org/mor-methods-howto/mor_methods.html Fixes #1563.

Currently we are stuck with implicit/explicit Euler. It should be easy to add more interesting methods.

See scipy.sparse.linalg.lgmres

- **Breaking change:** The `output_error_estimate_return_vector` argument is removed from `compute`/`estimate_output_error`/`output`. The behavior can now by configured when instantiating the reductor. Fixes #1671.

Replace 'pip-compile' by 'uv pip compile' for generating requirements files

5

Benefits: - much faster - cross-platform and cross-version - more helpful error messages - no more docker for `make ci_requirements` [uv](https://github.com/astral-sh/uv) is created by the same guys responsible for ruff.