feat: add gateway client with env-driven assembly

Includes config aggregation for multi-source env keys, from_env and
from_settings factories with explicit shared-backend injection,
gather_bounded, top-level exports, tightened import-linter layers with
the gate removed from the Makefile, and the finalized .env.example.

tests/unit/test_ports.py

@@ -31,48 +31,89 @@ class _DummyPermit:
 
 
 class _DummyLimiter:
-    async def try_acquire(self, source_key: str, est_tokens: int): return _DummyPermit()
-    async def acquire(self, source_key: str, est_tokens: int): return _DummyPermit()
-    async def source_stats(self, source_key: str): return SourceStats(0, 0, 0)
+    async def try_acquire(self, source_key: str, est_tokens: int):
+        return _DummyPermit()
+
+    async def acquire(self, source_key: str, est_tokens: int):
+        return _DummyPermit()
+
+    async def source_stats(self, source_key: str):
+        return SourceStats(0, 0, 0)
+
     async def mark_progress(self) -> None: ...
-    async def progress_age_s(self) -> float: return 0.0
+    async def progress_age_s(self) -> float:
+        return 0.0
 
 
 class _DummyGate:
-    async def try_enter(self, source_name: str, owner: str): raise NotImplementedError
-    async def record_success(self, entry): raise NotImplementedError
-    async def record_failure(self, entry, reason: str, force_open: bool): raise NotImplementedError
-    async def release_probe(self, entry): raise NotImplementedError
-    async def retry_after_s(self, sources): return 0.0
+    async def try_enter(self, source_name: str, owner: str):
+        raise NotImplementedError
+
+    async def record_success(self, entry):
+        raise NotImplementedError
+
+    async def record_failure(self, entry, reason: str, force_open: bool):
+        raise NotImplementedError
+
+    async def release_probe(self, entry):
+        raise NotImplementedError
+
+    async def retry_after_s(self, sources):
+        return 0.0
 
 
 class _DummyMw:
-    async def __call__(self, request, call_next): return await call_next(request)
+    async def __call__(self, request, call_next):
+        return await call_next(request)
 
 
 class _DummyTransport:
-    async def complete(self, *, messages, source, stream, overlay, call_id): raise NotImplementedError
+    async def complete(self, *, messages, source, stream, overlay, call_id):
+        raise NotImplementedError
 
 
 class _DummyCache:
-    async def get(self, key: str): return None
+    async def get(self, key: str):
+        return None
+
     async def set(self, key: str, value: str, ttl_s: int) -> None: ...
 
 
 class _DummySelector:
-    def order(self, sources, stats): return list(sources)
+    def order(self, sources, stats):
+        return list(sources)
 
 
 class _DummyStrategy:
-    def request_overlay(self, schema): return {}
-    def parse(self, text: str) -> Any: return {}
+    def request_overlay(self, schema):
+        return {}
+
+    def parse(self, text: str) -> Any:
+        return {}
 
 
 class _DummyRecorder:
     async def record_llm_call(
-        self, *, call_id, parent_call_id, session_id, model, provider, source_name,
-        messages, response, thinking, prompt_tokens, completion_tokens, usage_source,
-        latency_ms, ttft_ms, max_inter_token_ms, cache_hit, error, cost,
+        self,
+        *,
+        call_id,
+        parent_call_id,
+        session_id,
+        model,
+        provider,
+        source_name,
+        messages,
+        response,
+        thinking,
+        prompt_tokens,
+        completion_tokens,
+        usage_source,
+        latency_ms,
+        ttft_ms,
+        max_inter_token_ms,
+        cache_hit,
+        error,
+        cost,
     ) -> None: ...
 
 
@@ -95,10 +136,15 @@ def test_protocols_are_runtime_checkable(impl, protocol):
 
 
 def _decision(**overrides) -> GateDecision:
-    base = dict(
-        source_name="qwen_1", allowed=True, state=GateState.CLOSED,
-        epoch=0, is_probe=False, probe_owner=None, retry_after_s=0.0,
-    )
+    base = {
+        "source_name": "qwen_1",
+        "allowed": True,
+        "state": GateState.CLOSED,
+        "epoch": 0,
+        "is_probe": False,
+        "probe_owner": None,
+        "retry_after_s": 0.0,
+    }
     base.update(overrides)
     return GateDecision(**base)
 
@@ -141,9 +187,15 @@ class TestGateDecisionInvariants:
 
 class TestGateUpdate:
     def test_bounds(self):
-        u = GateUpdate(applied=True, state=GateState.CLOSED, epoch=0, failure_count=0, retry_after_s=0.0)
+        u = GateUpdate(
+            applied=True, state=GateState.CLOSED, epoch=0, failure_count=0, retry_after_s=0.0
+        )
         assert u.applied
         with pytest.raises(ValueError):
-            GateUpdate(applied=True, state=GateState.CLOSED, epoch=-1, failure_count=0, retry_after_s=0.0)
+            GateUpdate(
+                applied=True, state=GateState.CLOSED, epoch=-1, failure_count=0, retry_after_s=0.0
+            )
         with pytest.raises(ValueError):
-            GateUpdate(applied=True, state=GateState.CLOSED, epoch=0, failure_count=-1, retry_after_s=0.0)
+            GateUpdate(
+                applied=True, state=GateState.CLOSED, epoch=0, failure_count=-1, retry_after_s=0.0
+            )