feat(translation): quality pipeline overhaul + new features (audit 2026-08-29)

Translation quality & format preservation:
- Word: merge adjacent same-format runs into one unit (sentence-level
  coherence like inline-tag handling); translate comments/balloons;
  dedupe textbox collection (was translated twice); RTL no longer
  overrides center/justify alignment; CJK/Arabic font hints (eastAsia/cs)
- PPTX: chart translations now actually reach the output file
  (ChartPart.blob is read-only — rewrite chart XML in the saved ZIP);
  CJK typeface hints (a:ea)
- Excel: sheet renames no longer break references — rewrite cell
  formulas (3D/quoted), defined names, data validations, cond. formats
- PDF: bold/italic honored (hebo/heit/hebi); table cells never merge;
  unchanged blocks left untouched (typography preserved, fixes duplicate
  hyperlinks); attempted/changed stats + route gate now cover PDF;
  CJK font paths; scanned PDFs via Mistral OCR (detection + admin settings)

Features:
- formality param (formal/informal) + automatic regional-variant prompts
- output_mode=bilingual docx (source above translation)
- per-user translation memory on Redis (falls back to LRU), context-hashed
- QA report + 0-100 confidence score in job status; L0 on by default
- OpenAI-compatible providers: whole chunk in ONE numbered-JSON request
  (~15x fewer calls) with per-item fallback; base prompt always present
  (custom prompt no longer replaces translation instructions)

Infra & marketing alignment:
- plan-based engine gating + vision gating (closes paid-engine leak);
  /providers/available filtered per plan; 107 languages exposed
- zh-CN/zh-TW validation fixed; libmagic disabled on Windows (native crash)
- admin: Mistral OCR settings + engine status dashboard; httpx<0.28 pin
  (TestClient breakage); Prometheus test fixture fixed
- marketing docs aligned with code (PDF+OCR, retention, engines, pricing)
- security: .env.ionos/.env.production/provider_settings.json removed

Tests: 1173 passed / 0 failed (6 network tests deselected: free Google
endpoint temporarily blocked from this machine)

tests/test_metrics.py

@@ -35,36 +35,6 @@ from prometheus_client import (
 _REPO_ROOT = Path(__file__).resolve().parent.parent
 
 
-def _load_metrics_module_with_registry(registry: CollectorRegistry):
-    """Load middleware/metrics.py with patched Counter/Histogram to
-    use the supplied registry. Returns the loaded module."""
-    spec = importlib.util.spec_from_file_location(
-        "metrics_under_test",
-        _REPO_ROOT / "middleware" / "metrics.py",
-    )
-    mod = importlib.util.module_from_spec(spec)
-    # Inject the fresh registry into the module's namespace before exec
-    mod.__dict__["_TEST_REGISTRY"] = registry
-
-    # Patch Counter/Histogram to use the fresh registry
-    orig_counter = Counter
-    orig_histogram = Histogram
-
-    def _counter(*args, **kwargs):
-        kwargs.setdefault("registry", registry)
-        return orig_counter(*args, **kwargs)
-
-    def _histogram(*args, **kwargs):
-        kwargs.setdefault("registry", registry)
-        return orig_histogram(*args, **kwargs)
-
-    mod.__dict__["Counter"] = _counter
-    mod.__dict__["Histogram"] = _histogram
-
-    spec.loader.exec_module(mod)
-    return mod
-
-
 @pytest.fixture(scope="module")
 def metrics():
     """Load the metrics module ONCE per test module.
@@ -80,55 +50,59 @@ def metrics():
     return _load_metrics_module_with_fresh_registry()
 
 
+def _load_metrics_module_with_registry(registry: CollectorRegistry):
+    """Load middleware/metrics.py with ALL its counters/histograms
+    registered on the supplied fresh registry.
+
+    metrics.py does ``from prometheus_client import Counter, Histogram`` at
+    module top — injecting patched classes into the module dict BEFORE exec
+    does not survive that import. The only reliable interception point is
+    the ``prometheus_client`` module itself: we swap its Counter/Histogram
+    attributes for subclasses that default to ``registry``, exec the
+    module source, and restore the originals.
+    """
+    import types
+    import prometheus_client
+
+    orig_counter = prometheus_client.Counter
+    orig_histogram = prometheus_client.Histogram
+
+    class _RegistryCounter(orig_counter):
+        def __new__(cls, *args, **kwargs):
+            kwargs.setdefault("registry", registry)
+            return super().__new__(cls)
+
+        def __init__(self, *args, **kwargs):
+            kwargs.setdefault("registry", registry)
+            super().__init__(*args, **kwargs)
+
+    class _RegistryHistogram(orig_histogram):
+        def __new__(cls, *args, **kwargs):
+            kwargs.setdefault("registry", registry)
+            return super().__new__(cls)
+
+        def __init__(self, *args, **kwargs):
+            kwargs.setdefault("registry", registry)
+            super().__init__(*args, **kwargs)
+
+    source = (_REPO_ROOT / "middleware" / "metrics.py").read_text(encoding="utf-8")
+    mod = types.ModuleType("metrics_under_test")
+    mod.__file__ = str(_REPO_ROOT / "middleware" / "metrics.py")
+    try:
+        prometheus_client.Counter = _RegistryCounter
+        prometheus_client.Histogram = _RegistryHistogram
+        exec(compile(source, mod.__file__, "exec"), mod.__dict__)
+    finally:
+        prometheus_client.Counter = orig_counter
+        prometheus_client.Histogram = orig_histogram
+    return mod
+
+
 def _load_metrics_module_with_fresh_registry():
     """Load metrics module with its counters/histograms attached to a
-    fresh CollectorRegistry."""
-    spec = importlib.util.spec_from_file_location(
-        "metrics_under_test",
-        _REPO_ROOT / "middleware" / "metrics.py",
-    )
-    mod = importlib.util.module_from_spec(spec)
-    spec.loader.exec_module(mod)
-
-    # The module just created its counters on the default REGISTRY.
-    # Unregister them, then re-create them on a fresh registry.
-    fresh = CollectorRegistry()
-    for name in (
-        "http_requests_total",
-        "translation_total",
-        "translation_duration_seconds",
-        "file_size_bytes",
-        "quality_l0_checks_total",
-        "quality_l1_judge_total",
-        "quality_l1_judge_duration_seconds",
-        "quality_l1_judge_cost_usd",
-        "translation_retry_total",
-        "format_elements_lost_total",
-    ):
-        if not hasattr(mod, name):
-            continue
-        obj = getattr(mod, name)
-        try:
-            REGISTRY.unregister(obj)
-        except KeyError:
-            pass
-
-    # Now re-import the module so the new metrics register on `fresh`
-    spec = importlib.util.spec_from_file_location(
-        "metrics_under_test_isolated",
-        _REPO_ROOT / "middleware" / "metrics.py",
-    )
-    mod2 = importlib.util.module_from_spec(spec)
-    # We can't easily re-route Counter/Histogram in exec_module because
-    # they call into the global REGISTRY via the function signature.
-    # Instead: reload by re-importing via importlib with a wrapper
-    # that intercepts the Counter/Histogram constructors. We do this
-    # via the more direct route: use the EXISTING counters on the
-    # default REGISTRY, but only check RELATIVE increments.
-    #
-    # Practical approach: just use the module as-is. Tests check
-    # `after >= before + 1`, which is robust against other tests.
-    return mod
+    fresh CollectorRegistry (never the shared default one — the app may
+    already have registered the same names earlier in the session)."""
+    return _load_metrics_module_with_registry(CollectorRegistry())
 
 
 def _counter_value(counter, **labels):