fix: close color painting gizmo before batch match apply

[wudi901]

Description

Batch color-mix match rewrites painted extruder ids directly in the model (apply_batch_match_to_model + cleanup_unused_filaments_after_batch_match). While the Color Painting gizmo is open it keeps its own editing copy of that painting data in m_triangle_selectors, and its data_changed() only reloads the selectors when the extruder count changes — a same-count palette rewrite takes the colors-only branch and keeps stale facet→extruder assignments. Two consequences:

• with the painting panel open, the applied match rendered wrong (gizmo renders from its stale copy), and
• any subsequent stroke/gap-fill/erase-all committed via update_model_object() wrote the stale copy back over the applied match (durable corruption).

Fix: after the match apply finishes (model is final), force a re-deserialize of the gizmo's editing copy from the model via the new GLGizmoMmuSegmentation::refresh_from_model() (init_extruders_data() + init_model_triangle_selectors() — the same reload pairing data_changed() already uses on count changes). The panel stays open and the user's selected filament is preserved when possible.

Both canvases are covered (3D view and assemble view each own a GLGizmosManager instance); the call is a no-op where the gizmo is not active.

Screenshots/Recordings/Graphs

UI behavior change only (open painting panel now re-syncs in place instead of showing stale data); no visual layout change to capture.

Tests

Manual verification steps (wx UI event-loop path, not unit-testable):

1. Open the Color Painting gizmo, paint multiple colors, then run Color Mixing Match and confirm the match.
   • Expected: the painting panel stays open and the 3D view immediately shows the matched palette/mapping (no stale colors).
2. With the panel still open, paint one more stroke, then close/reopen the gizmo.
   • Expected: the stroke applies on top of the matched mapping and survives (no stale write-back).
3. Regression: run Color Mixing Match with the painting gizmo closed.
   • Expected: behavior unchanged.
4. Repeat step 1 in the assemble view.
   • Expected: same correct re-sync on the assemble canvas.

[wudi901]

Self-review (REQUEST_CHANGES-equivalent — 1 blocking finding)

Scope: 1 commit, +13 in src/slic3r/GUI/Plater.cpp (base build_platform_engineering, 0 behind base).

Verified correct:

• Placement: after result.success, before any palette/filament/painting mutation — gizmo cache is dropped before the model rewrite.
• reset_all_states() on gizmo Off clears m_triangle_selectors without writing back — no new overwrite risk.
• API usage mirrors existing compiled call sites (Plater.cpp ~L16900 same chain; GUI_ObjectList.cpp ~L1457 same manager pattern); 140-col limit respected; comment explains WHY.

Blocking (P2) — see inline comment: guard only checks the 3D-view canvas manager. Each canvas owns its own GLGizmosManager; painter gizmos support the assemble view (CanvasAssembleView branches in GLGizmoPainterBase.cpp:100/147/312; accessor Plater::get_assmeble_canvas3D() at Plater.hpp:693). If Color Painting is active on the assemble canvas, the stale-cache bug survives. Reset both managers — reset_all_states() is a no-op when nothing is active there.

Process note: no CI ran for this fork PR (0 check-runs on head SHA); full Windows build not run in the review sandbox. Compile risk minimal (mirrored patterns), but tick CI / local build_release_vs2022.bat before merge.

[wudi901]

[P2] Assemble-view painting gizmo is not covered by this guard.

Only get_view3D_canvas3D()->get_gizmos_manager() is checked, but every canvas owns its own GLGizmosManager and painter gizmos explicitly support the assemble view (CanvasAssembleView branches in src/slic3r/GUI/Gizmos/GLGizmoPainterBase.cpp:100/147/312, accessor Plater::get_assmeble_canvas3D() at src/slic3r/GUI/Plater.hpp:693). If Color Painting is open on the assemble canvas when the match is applied, its m_triangle_selectors cache stays stale — the exact bug this PR fixes, on the other canvas.

Suggested fix (no-op when nothing is active on that manager):

Plater* batch_plater = wxGetApp().plater();
for (GLCanvas3D* cnv : { batch_plater->get_view3D_canvas3D(), batch_plater->get_assmeble_canvas3D() }) {
    if (cnv == nullptr) continue;
    GLGizmosManager& gizmos_mgr = cnv->get_gizmos_manager();
    if (gizmos_mgr.get_current_type() == GLGizmosManager::EType::MmSegmentation)
        gizmos_mgr.reset_all_states();
}

[wudi901]

Addressed in e96f1c9 — approach switched from closing the gizmo to re-deserializing its editing copy from the final model state:

• New public GLGizmoMmuSegmentation::refresh_from_model() calls init_extruders_data() + init_model_triangle_selectors() — the same reload pairing data_changed() already uses on count changes, but unconditional, so same-count rewrites also re-sync.
• Called after cleanup_unused_filaments_after_batch_match() (model final), before the final canvas/panel refreshes.
• Both canvases covered (3D view + assemble view each own a GLGizmosManager), no-op where the gizmo isn't active — this also fixes the assemble-view gap raised here.
• UX preserved: the painting panel stays open; the selected filament is kept when still present.