feat(web): add right-click context menu and search/navigation to Web GUI

[jorge-ferreira-pii]

Addresses a portion of #10619

2.7 Right-click context menu

Feature	GUI	Web	Status
Select → Connected (insts / output / input / all nets / buffer trees)	LayoutViewer::selectHighlightConnectedInst/Nets/BufferTrees → Gui::selectHighlightConnected*	right-click first selects the object under the cursor, then a context_action (select_connected_insts / select_connected_output+input nets / select_connected_nets / select_connected_buffer_trees); the server handleContextAction builds the set once via computeConnectedSet. The buffer-tree walk can toggle whether inverters are followed, and the number of objects added is surfaced as a toast	❌ → ✅
Menu enabled by selection type	LayoutViewer::updateContextMenuItems() + Gui::anyObjectInSet() disabling menu_actions_[kSelect*Act]	every selection response is tagged with sel_has_inst / sel_has_net (addSelectionTypeFlags); the menu greys out items that don't apply — e.g. over a net only Connected Insts is enabled, and both submenus are disabled with nothing selected	❌ → ✅
Highlight → same + color groups	connected-* functions with a highlight_group, colored by Painter::kHighlight	the Highlight submenu issues the same context_actions with a group index and a 4×4 color-picker; overlay shapes (rects and flight lines for unrouted nets) reuse the 2.6 highlight pipeline (16 groups)	❌ → ✅
View → Zoom In / Out / Fit	LayoutViewer::zoomIn / zoomOut, Gui::zoomTo	the View submenu is dropped from the context menu; Fit becomes an on-canvas L.Control next to Leaflet's built-in zoom ± buttons (plus the f shortcut), where users expect it	❌ → ✅
Save (visible / entire)	Gui::saveImage / gui::save_image (server-side render)	client-side WYSIWYG composite (capture.js) of the displayed scene — base tiles + heatmap + highlight/selection overlay + ruler/label SVG panes — drawn onto a canvas via per-tile getBoundingClientRect (correct under fractional zoom); Visible captures the viewport, Entire fits the whole design, waits for the tiles to settle, captures, then restores the previous view	❌ → ✅
Clear (selections / focus nets / route guides / rulers / all)	Gui::clearSelections / clearHighlights / clearRulers	context_action clear_selections / clear_focus_nets / clear_route_guides / clear_all; rulers are cleared client-side	❌ → ✅

2.8 Search & navigation

Feature	GUI	Web	Status
Find (by name / glob)	MainWindow::showFindDialog → FindObjectDialog → block->findInst / findNet	new WebSocketRequest::kFind + SelectHandler::handleFind: {obj_type: inst|net|port, pattern, match_case}, matched by glob (fnmatch, optional case-fold), selection capped at 1000; a modal (type dropdown + Match Case + close button) drives it. Web addition: the view auto-zooms to the union bbox of the matches (map.fitBounds) — the Qt dialog only selects — and the hit count is reported	❌ → ✅
Go to position (x, y, size)	MainWindow::showGotoDialog → GotoLocationDialog	frontend-only modal (µm, prefilled with the current view center); converts µm→DBU and map.setView, or fitBounds a box of the requested size; Shift+G shortcut	❌ → ✅

[gemini-code-assist]

Code Review

This pull request introduces several new features to the web-based layout viewer, mirroring capabilities from the Qt GUI. These additions include a canvas right-click context menu for selecting and highlighting connected objects, search and navigation dialogs (Find and Go to Position), and a WYSIWYG layout capture tool that composites Leaflet layers and SVG overlays into a downloadable PNG. On the backend, support has been added for handling find and context action requests, rendering colored polygons, and serving image downloads via a new HTTP route. The review feedback highlights a few important issues: an uninitialized Color bg struct in web.cpp that could lead to undefined behavior, a logic flaw in capture.js where transparent background values prevent the fallback color from being applied, and a duplicate hex color code for dark_magenta in context-menu.js.

[gemini-code-assist]

The Color bg; variable is declared on the stack without initialization. In C++, POD structs are not zero-initialized by default when declared on the stack, which leads to undefined behavior (reading garbage values from the stack for the background color of the exported image). Value-initializing it with {} ensures it is safely zero-initialized (transparent).

Suggested change

	Color bg; // {0,0,0,0}
	Color bg{}; // {0,0,0,0}

[gemini-code-assist]

In JavaScript, standard transparent background values like 'transparent' or 'rgba(0, 0, 0, 0)' are truthy strings. Therefore, the logical OR fallback || '#111' will never be triggered if the computed background is transparent. This results in a transparent background in the exported PNG, making light-colored layout elements invisible on standard white image viewers. Checking for these transparent values explicitly ensures the fallback background color is correctly applied.

Suggested change

	ctx.fillStyle = getComputedStyle(container).backgroundColor || '#111';
	ctx.fillStyle = (getComputedStyle(container).backgroundColor !== 'transparent' && getComputedStyle(container).backgroundColor !== 'rgba(0, 0, 0, 0)') ? (getComputedStyle(container).backgroundColor || '#111') : '#111';

[gemini-code-assist]

The hex color for dark_magenta is set to #800080, which is identical to the hex color for purple (also #800080). Standard dark magenta is #8b008b. Correcting this hex value avoids duplicate colors and ensures the color swatches in the context menu are distinct.

Suggested change

	{ name: 'dark_magenta', hex: '#800080' },
	{ name: 'dark_magenta', hex: '#8b008b' },