Expand README with activation function reference and model-to-activation mapping

[Copilot]

The issue asks for coverage of neural-network activation functions and how they relate to different model types. This PR turns the repository into a concise, scannable reference for both.

• Activation function coverage

  • Added core activations: Sigmoid, Tanh, ReLU, Leaky ReLU, ELU, and Softmax.
  • For each: included formula, output range, and practical use case.

• Model-oriented guidance

  • Added typical activation pairings for MLPs, CNNs, RNN/LSTM/GRU, and Transformers.
  • Included a quick selection guide for common task patterns (binary vs multi-class outputs, ReLU fallback choices).

• Documentation structure

  • Reworked README.md from title-only content into a compact reference layout for quick lookup.

### 3) ReLU
- Formula: `ReLU(x) = max(0, x)`
- Output range: `[0, ∞)`
- Common use: default hidden-layer activation in many deep networks.

Original prompt

Working about various activation functions of NN and different models