This PR adds a test case to psbt2_timelock_test covering an unsatisfiable locktime combination in PartiallySignedTransaction::ComputeTimeLock().
When different PSBT v2 inputs specify their own timelock requirement, ComputeTimeLock() needs to reconcile all of those into a single locktime for the whole transaction. To reconcile this locktime, all the inputs locktimes need to be height or time-based, but not a mix of them.
The existing test already covers this failure when the input #0 is height-based and a later input is time-based, returning std::nullopt.
This PR adds the other case when the input #0 is time-based and a later input is height-based, returning std::nullopt.
I've basically swapped the PSBT inputs from the already existing case to cover this one.