docs: reproducible RESPACK (SrVO3) tutorial

[k-yoshimi]

Summary

Add a "RESPACK workflow" section to the cif2x tutorial (Japanese and English), following the design doc committed on this branch (docs/superpowers/specs/2026-07-01-respack-tutorial-design.md). A reader can download the pinned pseudopotential set, run cif2x -t respack input.yaml SrVO3.cif, and obtain exactly the committed scf/scf.in / nscf/nscf.in / input.in.

Stacked on #16 (the respack sample-shape fix); the base is set to that branch so this PR shows only the tutorial changes, and it will retarget to main automatically when #16 merges.

Changes

• docs/tutorial/cif2x/respack/ fixtures: primitive cubic SrVO3.cif, 3-task input.yaml (scf / nscf / respack), namelist-only respack.in_tmpl (N_wannier = 3 for the V t2g manifold, cRPA on), pp.csv / cutoff.csv for the pinned SG15 ONCV v1.0 set, and a README with download instructions.
• Pseudopotentials are not bundled: the tutorial gives the exact curl URLs and the rename step to cif2x's <element>.<name>.UPF convention. Cutoffs (ecutwfc 80 Ry / ecutrho 320 Ry) are supplied via optional.cutoff_file because ONCV UPF headers carry no usable cutoff metadata.
• Committed cif2x outputs generated with the real SG15 UPFs, shown via literalinclude.
• New tutorial section appended to docs/{ja,en}/source/cif2x/tutorial/index.rst: workflow overview, input parameter file, RESPACK template (user physics vs auto-generated k-path/SCDM), pseudopotential preparation, generation, and the downstream commands (pw.x → qe2respack → calc_wannier/chiqw/w3d/j3d) — commands only, no physics results.

Verification

• make html succeeds for both docs/ja and docs/en; all literalinclude paths resolve. The only new warning is the same benign fortran-lexing retry ({automatic} token) that the existing QE tutorial's scf.in include already emits.
• Regenerating the outputs from the committed fixtures is deterministic (byte-identical diff).
• pytest tests/ → 268 passed.

🤖 Generated with Claude Code