Standardize milestone naming with numbered sequence and historical anchors

Applied consistent naming pattern: 0X_[figure]_[task].py

M01 (1957 Perceptron):
- forward_pass.py → 01_rosenblatt_forward.py
- perceptron_trained.py → 02_rosenblatt_trained.py

M02 (1969 XOR):
- xor_crisis.py → 01_xor_crisis.py
- xor_solved.py → 02_xor_solved.py

M03 (1986 MLP):
- mlp_digits.py → 01_rumelhart_tinydigits.py
- mlp_mnist.py → 02_rumelhart_mnist.py

M04 (1998 CNN):
- cnn_digits.py → 01_lecun_tinydigits.py
- lecun_cifar10.py → 02_lecun_cifar10.py

M05 (2017 Transformer):
- vaswani_chatgpt.py → 01_vaswani_generation.py
- vaswani_copilot.py → 02_vaswani_dialogue.py
- profile_kv_cache.py → 03_vaswani_profile.py

Benefits:
- Clear execution order (01, 02, 03)
- Historical context (rosenblatt, lecun, vaswani)
- Descriptive purpose (generation, dialogue, profile)
- Consistent structure across all milestones

Updated documentation:
- README.md: Updated all milestone examples
- site/chapters/milestones.md: Updated bash commands

README.md

@@ -325,9 +325,9 @@ python xor_solved.py
 
 ### 🔢 03. MLP Revival (1986) - After Module 07
 ```bash
-cd milestones/03_1986_mlp_revival
-python mlp_digits.py     # 8x8 digit classification
-python mlp_mnist.py      # Full MNIST dataset
+cd milestones/03_1986_mlp
+python 01_rumelhart_tinydigits.py     # 8x8 digit classification
+python 02_rumelhart_mnist.py          # Full MNIST dataset
 # Backpropagation revolution on real vision!
 # YOUR training loops achieve 95%+ accuracy
 ```
@@ -338,9 +338,9 @@ python mlp_mnist.py      # Full MNIST dataset
 
 ### 🖼️ 04. CNN Revolution (1998) - After Module 09
 ```bash
-cd milestones/04_1998_cnn_revolution
-python cnn_digits.py     # Spatial features on digits
-python lecun_cifar10.py  # Natural images (CIFAR-10)
+cd milestones/04_1998_cnn
+python 01_lecun_tinydigits.py  # Spatial features on digits
+python 02_lecun_cifar10.py     # Natural images (CIFAR-10)
 # LeCun's CNNs achieve 75%+ on CIFAR-10!
 # YOUR Conv2d + MaxPool2d unlock spatial intelligence
 ```
@@ -351,8 +351,10 @@ python lecun_cifar10.py  # Natural images (CIFAR-10)
 
 ### 🤖 05. Transformer Era (2017) - After Module 13
 ```bash
-cd milestones/05_2017_transformer_era
-python vaswani_shakespeare.py
+cd milestones/05_2017_transformer
+python 01_vaswani_generation.py  # Text generation
+python 02_vaswani_dialogue.py    # Interactive chat
+python 03_vaswani_profile.py     # Optimization analysis
 # Attention mechanisms for language modeling
 # YOUR attention implementation generates text!
 ```

site/chapters/milestones.md

@@ -98,9 +98,9 @@ Images → Flatten → Linear → ReLU → Linear → ReLU → Linear → Classe
 - Architecture: Multi-layer feature learning
 
 ```bash
-cd milestones/03_1986_mlp_revival
-python mlp_digits.py      # 8x8 digits (quick)
-python mlp_mnist.py       # Full MNIST
+cd milestones/03_1986_mlp
+python 01_rumelhart_tinydigits.py  # 8x8 digits (quick)
+python 02_rumelhart_mnist.py       # Full MNIST
 ```
 
 **Expected Results**: 95%+ accuracy on MNIST
@@ -128,9 +128,9 @@ Images → Conv → ReLU → Pool → Conv → ReLU → Pool → Flatten → Lin
 - Architecture: Local connectivity + translation invariance
 
 ```bash
-cd milestones/04_1998_cnn_revolution
-python cnn_digits.py          # Spatial features on digits
-python lecun_cifar10.py       # CIFAR-10 @ 75%+ accuracy
+cd milestones/04_1998_cnn
+python 01_lecun_tinydigits.py  # Spatial features on digits
+python 02_lecun_cifar10.py     # CIFAR-10 @ 75%+ accuracy
 ```
 
 **Expected Results**: **75%+ accuracy on CIFAR-10** ✨