Fix fire hull climate scalar mapping to respect hull time layers

[ttuff]

Motivation

• The vase/hull renderer was producing an almost-uniform purple mesh because climate scalars were misaligned with hull time layers when the per-day series included buffer days or different indexing. This change targets the scalar→mesh alignment only to restore meaningful, time-structured coloring.

Description

• Replaced the earlier prefix/truncate approach with explicit layer-aware mapping by deriving layer days from hull.t_days_vert, computing n_layers, and ensuring n_vertices % n_layers == 0 before expanding scalars to vertices.
• Map hull layer offsets to calendar dates with hull.event.t0 + pd.to_timedelta(...) and reindex the summary.per_day_mean onto those layer dates, using ffill().bfill() when necessary, then expand day_vals with np.repeat(..., verts_per_layer) to produce per-vertex intensities.
• Added strict validation checks that raise on obvious mismatches and kept an informative ValueError if expanded scalar length does not match vertex count.
• Added a diagnostic scalar_debug_mode="slice" (colors by slice index) alongside the existing scalar_debug_mode="z", and enriched the fire_hull_scalar_debug printout to include scalar_dtype, unique_count, n_layers, and verts_per_layer for better pre-render inspection.
• Added focused unit tests to cover buffered-date alignment and the new slice diagnostic while preserving existing layer-order and z diagnostics (tests/test_fire_hull_viewer_scalars.py).

Testing

• Ran pytest tests/test_fire_hull_viewer_scalars.py -q and observed all tests passing (4 passed, 1 warning).
• Ran pytest tests/test_fire_plot_cube_first.py tests/test_fire_plot_loader_calls.py -q to validate integration with the fire_plot paths and observed all tests passing (11 passed, 1 warning).

---

Codex Task