Refactor shader build logic

[moven0831]

Introduction

this PR mainly contains two sections:

1. Refactor for shader build logic
2. Init UniFFI APIs for metal_msm using mopro v2

related issues:

• Optimize Workgroup and Thread Allocation for Metal’s GPU Architecture #10
• Update gpu-acceleration-app with latest mopro v0.2.0 integration #74

Refactor for shader build logic

• In Shaders precompile instead of compiling on the fly #83, it introduces shader pre-compile for our new metal MSM, with additional ~20% performance improvement for the 2^20 testcase.
• We enhanced shader compilation with MSM_COMPILE_ALL_SHADERS=1 flag for ci tests so that the metal libraries for MSM_ONLY (normal case) can be distinguished from the test build.

Init UniFFI APIs for metal_msm using mopro v2

• We simply create the example/ directory using mopro init
• Import mopro-msm as a module and build an API to run metal msm