chore: refactor Triton kernels for improved readability and performance

- Updated `causal_linear_attention_kernel` in `attention.py` to enhance code clarity by restructuring parameter lists and adding comments.
- Modified `rate_encoding_kernel` in `encoding.py` for better organization and readability, including adjustments to the random value generation logic.
- Refined `linear_lif_fwd_kernel` in `linear_lif.py` to improve the clarity of matrix operations and voltage updates, ensuring better maintainability.
- Enhanced `SpikingSelfAttention` class in `attention.py` to streamline state management and improve code readability.
- Improved `LinearLIF` class in `linear_lif.py` by refining initialization and state handling, ensuring better compatibility with Triton kernels.

Author: Khushiyant