Motivation

• Expose a new Puter GPT-5.4 Nano model in the app and surface it at the top of the model selection UI.
• Ensure the actual selected model name is propagated into various service/request paths (Cerebras streaming and the background service intent) so requests use the current model.
• Simplify and harden offline/shot-prompt handling and ensure build types and samples variant debuggable flags are explicit.

Description

• Added a new ModelOption entry PUTER_GPT_5_4_NANO with display name and provider set to PUTER and enabled screenshot support.
• Reordered the model dropdown to put PUTER_GPT_5_4_NANO before MISTRAL_MEDIUM_3_1 and excluded it from the normalModels filter so it appears in the ordered list.
• Updated the helper hint for PUTER_GLM5 to reference "GPT-5.4 Nano" punctuation.
• Changed Cerebras streaming request body construction to use the currently selected model name via GenerativeAiViewModelFactory.getCurrentModel().modelName instead of a fixed modelName variable.
• Updated the creation of the ScreenCaptureService intent to pass modelName = currentModel.modelName so the background service receives the selected model.
• Removed latestUserTaskInput usage and simplified createGenericScreenshotPrompt() to return an empty string (cleaning up stale task-input tracking).
• Made debug and release build types explicitly set isDebuggable, and ensured the samples build type is initialized from debug but explicitly marked non-debuggable.

Testing

• Ran a Gradle build locally with ./gradlew build and unit tests with ./gradlew test, and both completed successfully.

Codex Task