* for packets sent to ifu but not yet written back at replay,
  we still allow them to have predecode mispredict once written
  back from ifu

Fix Bim read idx bug
Remove valids in BranchPredictionResp
Modify out from Decoupled to Output in BasePredictorIO

* jumpOH could have more than one valid

* ifu: f2_redirect should consider successive predecode mispredicts
* ftq: when predicted not taken, cifIndex should be set to PrediectWidth-1

* inserted br should only add 1 new valid bit to brValids

* this target represents the target of the first jal instruction
  in the range of (FetchWidth * 4Bytes) regardless of whether it
  is in the real fetch stream

* save pftAddr for ifu req
* save nextRangeAddr(startAddr+FetchBytes) for pc read

This reverts commit 0a2b8431.

* FTB: fix getJmpOffset function
* Ftq: fix set_replay_status_between function
* IFU: toFtq.valid only holds 1 cycle

* pass the range vector into predecode to mask invalid instructions
* set the oversize bits when the new ftb entry has an RVI cfi at the end
* pass oversize bit to ifu to indicate this block has more than FetchWidth*4 Bytes