Design a better task scheduling system

[JiayinCao]

Some of the tasks, like shader compiling, is still not parallel. I need a better task scheduling system that does this

• Dynamic scheduling in a scheduled task during its execution.
• More flexible synchronization between tasks, maybe I need to introduce a 'barrier' concept.
• More abstraction in code.

[JiayinCao]

SORT switched to marl.

I still prefer to have my own job system eventually, so this is not closed. But the priority is low