feat: DeepSeek V4 thinking-mode round-trip via DeepSeekChatOpenAI subclass

Resolves #599: thinking-mode models require reasoning_content to be
echoed back across turns; multi-turn agent runs failed with HTTP 400.

The fix isolates DeepSeek's quirks (reasoning_content round-trip and
the deepseek-reasoner no-tool_choice limitation) into a subclass so
the general OpenAI-compatible client stays untouched. Adds DeepSeek
V4 Pro/Flash to the catalog. 9 new tests; rationale documented in
the class docstrings.

Design adapted from #600; #611 closed in favour of this approach.

tradingagents/llm_clients/model_catalog.py

@@ -65,10 +65,12 @@ MODEL_OPTIONS: ProviderModeOptions = {
     },
     "deepseek": {
         "quick": [
+            ("DeepSeek V4 Flash - Latest V4 fast model", "deepseek-v4-flash"),
             ("DeepSeek V3.2", "deepseek-chat"),
             ("Custom model ID", "custom"),
         ],
         "deep": [
+            ("DeepSeek V4 Pro - Latest V4 flagship model", "deepseek-v4-pro"),
             ("DeepSeek V3.2 (thinking)", "deepseek-reasoner"),
             ("DeepSeek V3.2", "deepseek-chat"),
             ("Custom model ID", "custom"),