Misbehaviour: wrong timestamp being used in trusting period checks

[plafer]

In checking for misbehaviour, we use host_timestamp() as the timestamp used in “trusting period checks”. This is not the timestamp we want to be using though, since this gives the timestamp associated with the latest “trusted consensus state”, whereas we want to be using the timestamp of the header we’re processing.

Equivalent in ibc-go (assuming ctx.BlockTime() is the height of the header being processed, which I think it is).

[plafer]

This is not a bug; I think this was a misunderstanding of mine while doing #584.