Inconsistent swap progress event emitting in state machine

[Einliterflasche]

Sometimes we emit the swap progress event of e.g. BtcRefunded before transitioning into that state, sometimes we emit it once inside that state (e.g. BtcCancelled). The second one is better because of less duplicated code incase we transition into a state from multiple other states.

We should also think about emitting the pure state-transition events which correspond very closely to the state machine from the function that calls next_state.