Feature/e2k support

[png-tech]

Overview

Add native e2k (Elbrus 2000) support to llama.cpp for CPU-based inference.

Motivation:

Enable efficient local inference on Elbrus-based systems (e.g., Elbrus‑8SV) using ggml-cpu backends, with attention to instruction compatibility, floating-point semantics, and memory access patterns specific to the e2k architecture.

Changes:

• Add e2k-specific detection and backend selection logic.
• Disable or replace x86-specific intrinsics (e.g., F16C, AVX, etc.) with portable or e2k-appropriate implementations.
• Adjust math and quantization paths to avoid relying on unsupported instructions.
• Verify correctness and performance with common quantized formats (Q4_K_M, Q5_K_M, Q8_0).

Benefits:

• Allows running llama.cpp natively on e2k without illegal instruction faults.
• Maintains acceptable inference speed on CPU-only Elbrus setups.
• Improves portability for users targeting domestic/specialized hardware.

Requirements

• I have read and agree with the contributing guidelines
• AI usage disclosure: YES

[ggml-gh-bot]

Hi @png-tech, thanks for your contribution!

Per our contribution guidelines, the automated PR checker found the following issue(s) that need your attention:

• PR Template not respected: Please respect the template when creating a new pull request. Make sure to fill out all required sections.

---

Please note that maintainers reserve the right to make final decisions on PRs. If you believe there is a mistake, please comment below.