[jafarbakhshaliyev]

Description

Adds a new SpikFGo model implementation for multivariate time-series forecasting. The model introduces a spiking Fourier graph-operator style architecture with spike-domain encoding, frequency-domain processing, and decoder support for forecasting tasks.

This PR also includes CPG-style spike positional encoding support, where the spike input is concatenated with CPG-PE and projected back through a Linear + BatchNorm + LIF pipeline.

Type of Change

• Bug fix
• New model implementation
• Dataset / data loader addition
• Documentation improvement
• Reproducibility / results addition
• Infrastructure / tooling

Checklist

• Code runs locally without errors
• No datasets, checkpoints, or large output files are committed
• README and/or MODEL_ZOO.md updated if a new model is added
• Citation for original paper added (in model file and MODEL_ZOO.md)
• PROTOCOL.md and RESULTS.md updated if new results are included
• Scripts under scripts/ are provided and reproduce the reported results
• If new dependencies are added, requirements.txt and environment.yml are updated

Related Issue

Closes #

Additional Notes

This PR currently focuses on adding the model implementation. Reproducibility scripts, model zoo entries, and benchmark results can be added either in this PR or a follow-up PR depending on repository expectations.