Pixels: Add unified_input_surface to unified input action pixels

[karlenDimla]

Task/Issue URL: https://app.asana.com/1/137249556945/task/1216706777618506?focus=true
Tech Design URL (if applicable): N/A
API Proposals URL(s) (if applicable): N/A

Description

Adds a surface parameter (unifiedInputSurface) to the contextual Unified Input UTI pixels, distinguishing where each pixel was fired from: address_bar, duck_ai, or contextual_chat.

• Introduces a DuckChatPixelSurface enum with a from(NativeInputState.InputContext) mapping (BROWSER → address_bar, DUCK_AI → duck_ai, DUCK_AI_CONTEXTUAL → contextual_chat).
• Threads the surface through all Unified Input pixel firing methods (tool select/deselect/submit, model & reasoning pickers, attachments, stop generation, customize responses) so callers pass the originating surface.
• Registers the new unifiedInputSurface param in params_dictionary.json and adds it to the affected m_aichat_unified_input_* pixel definitions in duck_chat.json5.
• Updates the ViewModels (attachments, native input mode widget, model/reasoning/options pickers) to supply the current surface when firing pixels.

Steps to test this PR

1. Address bar (surface=address_bar)

• Open the Unified Input from the address bar / omnibar
• Toggle a tool (web search / image generation), open the model picker, and attach/remove an image
• Each fired m_aichat_unified_input_* pixel includes surface=address_bar

2. Duck.ai tab (surface=duck_ai)

• Repeat the same interactions from the dedicated Duck.ai tab
• Each fired pixel includes surface=duck_ai

3. Contextual chat sheet (surface=contextual_chat)

• Open the contextual chat sheet over a web page
• Repeat the same interactions
• Each fired pixel includes surface=contextual_chat

---

Note

Low Risk
Analytics-only pixel schema and firing changes with no auth, data handling, or user-facing behavior changes.

Overview
Unified Input analytics now record where an interaction happened by adding a surface parameter (address_bar, duck_ai, or contextual_chat) to the relevant m_aichat_unified_input_* pixel definitions and wiring it through the app.

A DuckChatPixelSurface enum maps NativeInputState.InputContext to those values; DuckChatPixels APIs for tools, prompts, model/reasoning pickers, attachments, stop generation, customize responses, and FE recovery flows now require a surface and send it as surface on the wire. ViewModels derive the current surface from native input state when firing pixels. unifiedInputSurface is registered in params_dictionary.json. Unit tests were updated to expect the new parameter.

^{Reviewed by Cursor Bugbot for commit 3720d32. Bugbot is set up for automated code reviews on this repo. Configure here.}

[karlenDimla]

• Pixels: Add unified_input_surface to unified input action pixels #9306 👈 (View in Graphite)
• develop

This stack of pull requests is managed by Graphite. Learn more about stacking.

[github-actions]

Privacy Review task: https://app.asana.com/0/69071770703008/1216913713452942

[joshliebe]

Tested! "surface" sent as expected ✅