Fix prepare request in recovery message

[ZhangTao1596]

ValidatorIndex is uninitialized and unassigned in PrepareRequestMessage of RecoveryMessage. This will cause prepare request reverify failure when recovery.

[roman-khimov]

It can be set on the receiver side as well, somewhere in GetPrepareRequestPayload:

neo-modules/src/DBFTPlugin/Messages/RecoveryMessage/RecoveryMessage.cs

Lines 89 to 95 in c6d9640

	internal ExtensiblePayload GetPrepareRequestPayload(ConsensusContext context)
	{
	if (PrepareRequestMessage == null) return null;
	if (!PreparationMessages.TryGetValue(context.Block.PrimaryIndex, out PreparationPayloadCompact compact))
	return null;
	return context.CreatePayload(PrepareRequestMessage, compact.InvocationScript);
	}

That's what NeoGo currently does in https://github.com/nspcc-dev/neo-go/blob/5eb4ba772f4d0e507d8a77699b3856f340c36fe8/pkg/consensus/recovery_message.go#L205

[vncoelho]

I agree with @roman-khimov , this will also be better for mainnet compatibility at this moment.

[vncoelho]

@ZhangTao1596, I am double checking your commit, in fact, it will not affect payload serialization/de-serialization itself.
As we see, currently, it is not assigned and keeps null?

[vncoelho]

@superboyiii, if possible try an extended test.

We use to have Unit Tests for consensus, that we created sometime ago.
@erikzhang, do you know why they were not ported for neo-modules?

That Akka tests can cover several cases, such as this.

[superboyiii]

@superboyiii, if possible try an extended test.

We use to have Unit Tests for consensus, that we created sometime ago. @erikzhang, do you know why they were not ported for neo-modules?

That Akka tests can cover several cases, such as this.

I will.

[superboyiii]

Tested, it works well.