Backend: change ready logic from dispatch to decode for better performance

Backend: fix ready timing from dispatch to frontend (#3127)

Backend,util: add PipeGroupConnect to make group block in a better way

* PipeGroupConnect can block parts of data group in order. There is no need to handle part block in consumer's stage.
* Pipe re

Backend,util: add PipeGroupConnect to make group block in a better way

* PipeGroupConnect can block parts of data group in order. There is no need to handle part block in consumer's stage.
* Pipe regs will be refilled after the cycle when all data is accepted by next stage.
* There would be timing problem in canAcc signal, if the outReadys come too late. A possible implementation is providing configurable option whether canAcc refers outReadys, which will be provided when there is timing violation in this module.

show more ...