feat: wire model_capability into pipeline — tier-adaptive prompts + ctx auto-detect

Model tier detection now active (was orphaned code):
- orchestrator.__init__: detect tier from generator.llm.ollama_model
- ctx.model_tier written at run() start, flows to Generator
- Generator._build_system(): SMALL gets strict format constraints
  (1 function, ≤10 lines, preserve indent, no explanation, ≤15 words between tools)
  LARGE gets minimal constraint; MEDIUM uses base rules as-is
- _get_max_retries() respects model_strategy.max_retries

Context window auto-detection (get_effective_ctx):
- 4-layer probe: llama.cpp /props → vLLM /v1/models → Ollama modelinfo.llama.context_length → tier default
- Passed to LLMBackend n_ctx in build_pipeline
- SMALL=16K, MEDIUM=32K, LARGE=64K defaults; native ctx capped at 64K×0.8

501 tests green, 0 regression.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

kaiwu/cli/commands/task.py

@@ -71,10 +71,15 @@ def build_pipeline(model_path, ollama_url, ollama_model, project_root, verbose):
     _cfg = _load_cfg().get("default", {})
     _api_key = _cfg.get("api_key", "")
 
+    # Ctx自适应：检测模型可用上下文窗口
+    from kaiwu.core.model_capability import get_effective_ctx
+    effective_ctx = get_effective_ctx(ollama_model, ollama_url)
+
     llm = LLMBackend(
         model_path=model_path,
         ollama_url=ollama_url,
         ollama_model=ollama_model,
+        n_ctx=effective_ctx,
         verbose=verbose,
         api_key=_api_key,
     )