Feature: optimize compact performance improvement

[dantengsky]

Summary

Execution of Optimize table x compact statement is rather conservative, utilization of CPU / Bandwidth is rather low.

[BohuTANG]

compact blocks_select read segment files is not parallel, so it's slow:
https://github.com/datafuselabs/databend/blob/61589ffbef98db662b661a63a6c3f1565719dfa6/src/query/storages/fuse/src/operations/mutation/compact_mutator.rs#L95-L155

[BohuTANG]

Another improvement is to compact by batch instead read all the segments one by one compact.

[BohuTANG]

I am trying to read segments in parallel using read_segments