Improve performance of some opt passes

[annagrin]

Improve performance of some of the compiler passes

This change runs some passes on FuncOp instead of ModuleOp so the optimizations can run in parallel on multiple functions.

• Use cc-loop-unroll and cc-loop-normalize as func passes instead of module passes
• Make state-prep a func pass as well
• Update hardware pipelines
• Update tests

Related:
#2422
#2416