Evaluate integration test stability

[clangenb]

We have a trade-off between CI execeution time vs. Stability and the tuning parameter is the waiting-blocks after an extrinsic to ensure that our extrinsic is in the canonical chain.

Initially, it looked like 5 blocks of waiting is enough, but now we do get some failures due to tx reordering (i.e., next-phase accidentally before registering). Maybe it makes sense to increase the number to 7.

EDIT: Current CI fail rate seems to be around 25%

[clangenb]

As mentioned here, #234 (comment), I believe that we have a reasonable stability trade-off and we just hope that future substrate updates can make our CI more stable.