Complete EN coverage for 4 missing notebooks and restore I-probability emoji

- New EN translations: 16-function-calling, 18-mla-kv-cache,
  21-quantization, 23-inference-systems (all 1:1 with ZH: cell count,
  zero residual CJK, matching ✅ emoji)
- I-probability-information: restore ✅ in exercise prints to match ZH
- All 41 ZH notebooks now have a 1:1 EN counterpart at file level

Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>

Author: sanbuphy

notebooks-en/appendix-advanced/I-probability-information.ipynb

@@ -776,7 +776,7 @@
     "assert manual_H is not None, \"please compute manual_H first\"\n",
     "assert abs(manual_H - torch_H) < 1e-6, f\"answer should be {torch_H:.6f}, you got {manual_H}\"\n",
     "\n",
-    "print(f\"Exercise 1 passed:\")\n",
+    "print(f\"✅ Exercise 1 passed:\")\n",
     "print(f\"   H(p) = {manual_H:.4f}\")\n",
     "print(f\"   theoretical maximum log(4) = {math.log(4):.4f}\")\n",
     "print(f\"   p is more certain than uniform, so its entropy is smaller than log(4)\")\n"
@@ -819,7 +819,7 @@
     "assert torch.allclose(manual_lp, torch_lp, atol=1e-5), \\\n",
     "    f\"result inconsistent with PyTorch\\nyou got:    {manual_lp.tolist()}\\nPyTorch: {torch_lp.tolist()}\"\n",
     "\n",
-    "print(\"Exercise 2 passed:\")\n",
+    "print(\"✅ Exercise 2 passed:\")\n",
     "print(f\"   log_softmax = {[round(x, 4) for x in manual_lp.tolist()]}\")\n",
     "print(f\"   corresponding probs = {[round(x, 4) for x in torch.exp(manual_lp).tolist()]}\")\n",
     "print(f\"   sum of probs = {torch.exp(manual_lp).sum().item():.6f}\")\n",
@@ -867,7 +867,7 @@
     "assert abs(ce - expected_ce) < 1e-6\n",
     "assert abs(ppl - expected_ppl) < 1e-3\n",
     "\n",
-    "print(f\"Exercise 3 passed:\")\n",
+    "print(f\"✅ Exercise 3 passed:\")\n",
     "print(f\"   average CE = {ce:.4f}\")\n",
     "print(f\"   PPL       = {ppl:.4f}\")\n",
     "print(f\"   vocabulary V = {V}\")\n",