Test malicious-customer pay scenarios

[marsella]

We need to test scenarios with properly malicious customer, e.g. one who does not use our lovely, misuse-resistant API.

Consider the following scenarios:

• A malicious customer tries to pay on a closed channel. They form a pay request on a state that has already been closed (zeekoe does not stop them).
• A malicious customer tries to pay on a channel using an outdated state

In both these cases, the merchant must reject the payment with a ReusedRevocationLock error.

To implement this (and other malicious-party tests), we need to implement a party that adheres to the protocol flow but behaves incorrectly.

• Discuss possible architectures for implementing these tests

[indomitableSwan]

Just a quick note that we should make sure to test all close methods.