Split model and effort pickers

[jasonkeung]

Description

Splits noisy combined model/effort model pickers into separate base-model and effort-level controls.

• Adds shared grouped model/effort picker behavior for orchestration launch/config controls while keeping the selected value as the existing concrete combined model_id.
• Uses existing model metadata (base_model_name and reasoning_level) for Warp/Oz models, with conservative display-name grouping for harness model metadata.
• Updates onboarding's "Customize your Warp Agent" model selection to show separate Default model and Effort level dropdowns.
• Preserves backwards compatibility: no server/API/storage changes are required because the UI still emits the existing concrete model IDs.

Linked Issue

No GitHub issue; requested from Slack.

• The linked issue is labeled ready-to-spec or ready-to-implement.
• Where appropriate, screenshots or a short video of the implementation are included below (especially for user-visible or UI changes).

Testing

• cargo check --manifest-path /workspace/warp/Cargo.toml -p onboarding -p warp
• cargo test --manifest-path /workspace/warp/Cargo.toml -p warp orchestration_controls
• cargo fmt --manifest-path /workspace/warp/app/Cargo.toml --check
• cargo fmt --manifest-path /workspace/warp/crates/onboarding/Cargo.toml --check
• cargo build --manifest-path /workspace/warp/Cargo.toml -p warp --bin warp-oss
• I have manually tested my changes locally with ./script/run

Screenshots / Videos

Manual visual validation was done by launching /workspace/warp/target/debug/warp-oss and walking through onboarding. The "Customize your Warp Agent" screen showed separate Default model and Effort level controls; the model dropdown grouped base models, and the effort dropdown showed the valid effort choices for the selected base model.

I could not attach a saved screenshot from the sandbox because the available screenshot utilities/PIL dependencies were missing, but the UI was visually inspected through computer use after a successful build.

Agent Mode

• Warp Agent Mode - This PR was created via Warp's AI Agent Mode

CHANGELOG-OZ: Split Warp Agent model selection into separate model and effort level controls while preserving existing model IDs.

Conversation: https://staging.warp.dev/conversation/80bf636c-061f-46d5-9004-dbea4dcc3422
Run: https://oz.staging.warp.dev/runs/019e8b4d-585d-73ac-ab22-ca860bbc9b22
This PR was generated with Oz.