Add DiffractionSim and remove the abTEM dependency

[cophus]

What problem does it solve?

This PR adds DiffractionSim, an interactive electron diffraction simulator, and removes abTEM from quantem.widget.

DiffractionSim shows a crystal and its diffraction pattern side by side for any orientation, computed live with Bloch waves: nanobeam disks, CBED and Kikuchi patterns, with thickness, convergence, precession and energy controls. It was developed alongside the orientation and phase mapping work in quantEM (electronmicroscopy/quantem#297), whose Crystal and bloch modules it uses. A framework-free build of the same frontend (npm run build:web) runs on the group website.

abTEM was used only for Lobato scattering factors, one projected potential, cell orthogonalization and the interaction constant, but it brings in a large dependency stack including dask. We replaced these four pieces with quantEM's Lobato factors (Lobato and Van Dyck, Acta Cryst. A 70, 636 (2014), https://doi.org/10.1107/S205327331401643X) and short numpy implementations:

• ShowCIF potential previews and the scattering factors agree with abTEM to about 1e-5.
• PlanPtycho cells are identical to abTEM's, except for one case where abTEM's cell was missing an atom.
• The projected potential is now an exact Fourier sum. It differs from abTEM's by 0.5 to 4.5 % per pixel, because abTEM spreads each atom over four pixels; against an independent real-space reference the new potential is accurate to about 1e-4.
• The old cells also placed some atoms at -1e-17 A, which abTEM silently dropped, so SrTiO3 [111] and [112] potentials were 16 to 20 % too low. The new cells keep every coordinate in [0, 1).

The crystal extra is now ase>=3.23, spglib>=2.5 and pandas (pandas is used by PlanPtycho.report() and was previously supplied by abTEM). DiffractionSim, ShowCIF and PlanPtycho do not need quantem.gpu at runtime.

What should reviewer(s) do?

This is a draft until quantEM PR #297 is released. DiffractionSim, the ShowCIF potential preview and PlanPtycho now need quantem.diffraction from that PR. The quantem>= pin cannot be raised before the release, so their tests skip with that requirement, and the docs CI will fail on the PlanPtycho tutorial until quantEM is installed from #297.

Please review:

• DiffractionSim against the widget conventions: naming (diffractionsim), the export bridge, state_dict / save / load_state_dict, save_state with a frontend PNG fallback, validators, and the Title Case UI (src/quantem/widget/diffraction_sim.py, js/diffractionsim/).
• The abTEM replacements in src/quantem/widget/_lobato.py, _showcif_potential.py and planptycho.py, and the reference values in tests/test_showcif_potential.py and tests/planptycho/.

Verification

• tests/diffractionsim: 53 passed with quantEM from PR #297 (52 passed and 1 skipped without quantem.gpu installed). All DiffractionSim cases pass in tests/test_html_export_protocol.py and tests/test_save_state.py.
• ShowCIF and PlanPtycho tests: 81 passed in an environment where import abtem fails.
• abTEM 1.0.10 comparison (run in a throwaway environment): scattering factors to 1.2e-5, radial tables to 7e-6 of the peak, interaction constant to 6e-9, and identical oriented cells for 8 crystals and 11 zone axes each.
• npx vitest run js/diffractionsim and js/showcif: pass. tsc reports no new errors.
• JupyterLab: the widget renders; Export PNG and Export HTML work through the bridge ("Ready ... (1.1 MB, single file)"); the static fallback PNG reaches Python; Compute Reference renders the Kikuchi reference pattern.
• Not checked: Copy to the clipboard and the JupyterLab dark theme.

Generated with Claude Code

[github-actions]

• Keep cophus/quantem.widget:diffraction until this PR is merged.
• After merge, delete cophus/quantem.widget:diffraction if you own it and no other PR uses it.
• Then remove its dedicated worktree and local diffraction branch and update local main.

[bobleesj]

Either we can wait for architecture refactoring or merge now. up to you. probably I can do this during the weekend or so.