Add PyTorch/ONNX prediction models with physics fallback

- Train 3 MLP networks (acid speed 14→1, tension 4→10, quality 6→2)
  on 12,000 synthetic samples generated from physics models + noise
- Export pre-trained ONNX models to pt_models/ directory
- Rewrite prediction.py: ONNX inference first, physics fallback if unavailable
- Add onnxruntime + numpy to requirements.txt (Aliyun mirror for Docker)
- Use Tsinghua mirror in Dockerfile for pip installs

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

backend/app/services/pt_models/scalers.json

@@ -0,0 +1,118 @@
+{
+  "acid_speed": {
+    "X_mean": [
+      4.279637813568115,
+      11.041364669799805,
+      150.61114501953125,
+      149.8550262451172,
+      150.01771545410156,
+      150.00747680664062,
+      150.5503692626953,
+      149.680419921875,
+      69.46580505371094,
+      69.35639953613281,
+      69.35671997070312,
+      69.52980041503906,
+      69.55128479003906,
+      69.26476287841797
+    ],
+    "X_std": [
+      2.1389975547790527,
+      4.052000999450684,
+      51.74678421020508,
+      52.34047317504883,
+      51.52565383911133,
+      52.03013610839844,
+      51.94683074951172,
+      51.77798080444336,
+      10.097525596618652,
+      10.071165084838867,
+      10.012657165527344,
+      10.097661018371582,
+      10.077392578125,
+      10.07375717163086
+    ],
+    "y_mean": [
+      92.42212677001953
+    ],
+    "y_std": [
+      30.15455436706543
+    ]
+  },
+  "tension": {
+    "X_mean": [
+      4.211259365081787,
+      1104.45068359375,
+      376.4231262207031,
+      0.25053057074546814
+    ],
+    "X_std": [
+      2.1753084659576416,
+      290.6602783203125,
+      129.755859375,
+      0.057579852640628815
+    ],
+    "y_mean": [
+      439.6892395019531,
+      373.8921203613281,
+      342.94708251953125,
+      316.7646179199219,
+      298.9944763183594,
+      298.7807312011719,
+      307.93359375,
+      334.0721740722656,
+      387.0394592285156,
+      439.63824462890625
+    ],
+    "y_std": [
+      341.6143798828125,
+      290.6787109375,
+      266.5653991699219,
+      246.12326049804688,
+      232.24676513671875,
+      231.9054412841797,
+      239.40611267089844,
+      259.3957214355469,
+      300.75,
+      341.000244140625
+    ],
+    "zone_names": [
+      "inlet",
+      "s1_roller",
+      "acid_entry",
+      "acid1",
+      "acid2",
+      "acid3",
+      "rinse",
+      "leveler",
+      "s2_roller",
+      "outlet"
+    ]
+  },
+  "quality": {
+    "X_mean": [
+      4.245825290679932,
+      99.6146011352539,
+      150.18072509765625,
+      70.02484893798828,
+      10.967876434326172,
+      74.14215850830078
+    ],
+    "X_std": [
+      2.18487548828125,
+      46.58076858520508,
+      51.831016540527344,
+      11.553886413574219,
+      4.02312707901001,
+      31.69878578186035
+    ],
+    "y_mean": [
+      92.10986328125,
+      89.79142761230469
+    ],
+    "y_std": [
+      11.346452713012695,
+      8.92744255065918
+    ]
+  }
+}