docs(notebooks): run example notebooks under the v1 convention

[FBumann]

TODO (author): write your intent and motivation here in your own words.

Note

The following content was generated by AI.

Stacked on #717. Opts every example notebook into the v1 arithmetic convention (linopy.options["semantics"] = "v1") and fixes the code that relied on legacy alignment, so the docs teach and run under v1.

Notebooks needing more than the opt-in

• creating-variables — the transport mask was a raw (6, 6) numpy array; under v1 that can't be paired unambiguously with the (from, to) dims, so it's now a labelled DataArray.
• creating-expressions — the "first object determines the coordinates" positional example now raises. The notebook shows the ValueError and resolves it with join="outer"; the .loc[:4] + .loc[5:] positional trick became an explicit join="override".
• coordinate-alignment — the intro taught legacy superset-fill and same-shape positional alignment, both removed by v1. Rewrote it to teach label-based alignment (mismatches raise; resolve via .sel / .reindex / .assign_coords or an explicit join=). The join walkthrough is kept; the summary table now reflects v1 (default raises; outer/left/right fill non-overlapping slots with the operation's identity, verified against actual v1 output).

Remote notebooks (solve-on-oetc, solve-on-remote) carry the opt-in too. Both are nbsphinx: {execute: never}; their model-building code was verified v1-safe locally. solve-on-remote keeps its committed outputs (it's excluded from the nbstripout hook).

Verification

• All 13 locally-runnable notebooks execute cleanly under v1 (jupyter nbconvert --execute).
• Outputs stripped to match the nbstripout pre-commit hook; notebook metadata restored from the base so diffs are source-only.
• ruff, nbstripout, and the other pre-commit hooks pass.

🤖 Generated with Claude Code