fix(tui): restore auto model state on session load

[LING71671]

Summary

Closes #1797.

Resuming or explicitly loading a session saved while /model auto was active could restore the literal model string auto without restoring the matching auto_model state. The next turn could then treat auto as a concrete provider model and send it to the API.

Changes

• Route /load through App::set_model_selection so persisted model metadata rebuilds the full model-selection state.
• Clear stale effective auto-model/auto-reasoning labels when the model selection is restored.
• Restore thinking mode to Auto whenever the selected model is auto.
• Add regression coverage for explicit /load and apply_loaded_session resume paths.

Verification

• cargo fmt --all -- --check
• git diff --check
• cargo test -p deepseek-tui --bin deepseek-tui load_auto_model_session_restores_auto_mode
• cargo test -p deepseek-tui --bin deepseek-tui apply_loaded_session_restores_auto_model_mode
• cargo build

Note: cargo build still reports the existing unused migration marker warnings in schema_migration.rs; this change does not introduce new warnings.

[gemini-code-assist]

Code Review

This pull request updates the session loading mechanism to use the set_model_selection method, ensuring that application state—specifically reasoning_effort and last_effective_reasoning_effort—is correctly managed when restoring a session. It also introduces new test cases and updates existing ones to verify that the "auto" model selection mode is properly restored and that the internal state remains consistent. I have no feedback to provide.