refactor(ai_client): remove top-level SDK imports; use _require_warmed

Phase 3 T3.2 + T3.3 of startup_speedup_20260606 track.

The 5 heavy SDKs (anthropic, google.genai, openai, google.genai.types,
requests) are no longer imported at module level. Each function that
needs them now calls _require_warmed(name) to get the module from
sys.modules (populated by AppController's warmup on _io_pool).

This is the load-bearing wall of the Main Thread Purity Invariant:
heavy modules are never in the main thread's import chain.

run_discussion_compression now uses _require_warmed for both
google.genai.types (gemini branch) and requests (deepseek branch).

Tests/test_tier4_patch_generation.py adapted: the 2 tests that
mocked 'src.ai_client.types' (no longer a module-level attr)
now mock 'src.ai_client._require_warmed' (the new public mechanism).

T3.1 tests now pass (9/9). T3.3 breakage fixed.
All 25 ai_client + tier4 tests pass.

tests/test_tier4_patch_generation.py

@@ -36,31 +36,33 @@ def test_run_tier4_patch_generation_empty_error() -> None:
 
 def test_run_tier4_patch_generation_calls_ai() -> None:
   """Test that run_tier4_patch_generation calls the AI with the correct prompt."""
+  mock_types = MagicMock()
+  mock_types.GenerateContentConfig = MagicMock()
   with patch("src.ai_client._ensure_gemini_client"), \
      patch("src.ai_client._gemini_client", create=True) as mock_client, \
-     patch("src.ai_client.types") as mock_types:
+     patch("src.ai_client._require_warmed", return_value=mock_types):
     mock_resp = MagicMock()
     mock_resp.text = "--- a/test.py\n+++ b/test.py\n@@ -1 +1 @@\n-old\n+new"
     mock_client.models.generate_content.return_value = mock_resp
-    mock_types.GenerateContentConfig = MagicMock()
-    
+
     error = "TypeError: unsupported operand"
     file_context = "def foo():\n    pass"
     result = ai_client.run_tier4_patch_generation(error, file_context)
-    
+
     mock_client.models.generate_content.assert_called()
 
 def test_run_tier4_patch_generation_returns_diff() -> None:
   """Test that run_tier4_patch_generation returns diff text."""
+  mock_types = MagicMock()
+  mock_types.GenerateContentConfig = MagicMock()
   with patch("src.ai_client._ensure_gemini_client"), \
      patch("src.ai_client._gemini_client", create=True) as mock_client, \
-     patch("src.ai_client.types") as mock_types:
+     patch("src.ai_client._require_warmed", return_value=mock_types):
     expected_diff = "--- a/src/test.py\n+++ b/src/test.py\n@@ -10,5 +10,6 @@\n def test_func():\n-    old_value = 1\n+    old_value = 1\n+    new_value = 2"
     mock_resp = MagicMock()
     mock_resp.text = expected_diff
     mock_client.models.generate_content.return_value = mock_resp
-    mock_types.GenerateContentConfig = MagicMock()
-    
+
     result = ai_client.run_tier4_patch_generation("error", "context")
     assert "---" in result
     assert "+++" in result