test: try to mitigate flaky tests (#498)

* Add assertion to make sure all nodes are synced.

* Retry round event with BA's interval

We are about to modify lambdaDKG with block height, instead of timing.
Therefore, the last interval we could use is BA's interval.

* Check ctx in inifite loop

* Disable this panic temporary

If a fast sync triggered by receiving types.AgreementResult from later
position, this panic would be triggered 100%, ex.

current-round: 1,
change-notary-height: 120,
agreeemnt-result from round:2 height:121

However, I don't have easy way to fix it temporary.

* Lower down the test size for TCP

* Fix typo

* Fixup: revert the removed panic

1 files changed, 7 insertions, 0 deletions

diff --git a/core/agreement-mgr.go b/core/agreement-mgr.go
index 0e39fa5..14aa385 100644
--- a/core/agreement-mgr.go
+++ b/core/agreement-mgr.go
@@ -432,6 +432,13 @@ func (mgr *agreementMgr) baRoutineForOneRound(
 		var nextHeight uint64
 		var nextTime time.Time
 		for {
+			// Make sure we are stoppable.
+			select {
+			case <-mgr.ctx.Done():
+				breakLoop = true
+				return
+			default:
+			}
 			nextHeight, nextTime = mgr.bcModule.nextBlock()
 			if isStop(oldPos) && nextHeight == 0 {
 				break