feat: sync llama.rn

[a-ghorbani]

Description

This update syncs llama.rn (the pr is still open) : , which:

• Syncs llama.cpp with the b3972 release.
• Adds additional builds for the i8mm and sve libraries for Android.

Fixes: #73, #50, #25, #45 ((addresses multiple crash reports on Android)

Platform Affected

• [x] iOS
• [x] Android

Checklist

• [ ] Necessary comments have been made.
• [ x] I have tested this change on:
  • [x] iOS Simulator/Device
  • [x] Android Emulator/Device
• [x] Unit tests and integration tests pass locally.

[a-ghorbani]

I tested it on Pixel 9 Pro, Galaxy S24, and OnePlus 6 with the following models:

• Llama-3.2-1B-Instruct-Q8_0
• Llama-3.2-1B-Instruct-Q4_0_8_8
• Llama-3.2-1B-Instruct-Q4_0_4_8
• Llama-3.2-1B-Instruct-Q4_0_4_4

On these devices, all models worked fine except for Q4_0_4_4 on the OnePlus 6.