llama: use sliding window for phi3

[FanShupei]

• [x] I have read the contributing guidelines
• Self-reported review complexity:
  • [x] Low
  • [ ] Medium
  • [ ] High

Related issue report: #7709

This PR switches Phi3 model to use sliding window attention. After this PR, it no longer geneartes broken output after the 2,048 token. Tested on "phi3-mini-4k-instruct" model.

TODO: (DONE) ~~convert_hf_to_gguf.py changes~~