Implemented encoding null in key and value of a map in Protobuf

[shanshin]

Also improved null encoding error messages.

Resolves #2760

[shanshin]

The open question is whether we want to allow encoding null in the key.

This change may also make the encoded messages incompatible with some decoders created using protoc.

[sandwwraith]

The open question is whether we want to allow encoding null in the key.

How will it look like? A value without a key? In that case, we possibly can decode it automatically into a null: value map pair

[sandwwraith]

Is LIST_ELEMENT added for the purpose of proper error message only? I do not see this string being tested anywhere

[shanshin]

Yes, only the accepted or not flag is sufficient to throw an error. The enumeration is created only for printong different messages

[sandwwraith]

Then add a test for this message pls

[shanshin]

Tests added

[shanshin]

How will it look like? A value without a key?

Each map declaration in protobuf structually is repeated field with Entry message

message NullableMap {
    map<string, int32> m = 1;
}

is the same as

message NullableMap {
    message m_Entry {
        optional string key = 1;
        optional int32 value = 2;
    }

    repeated m_Entry m = 1;
}

see the protobuf doc

Therefore, we can skip either the key or the value.

In that case, we possibly can decode it automatically into a null: value map pair

It's already implemented in this PR, see encodedKeyNull in testNullMap.
Question: should we leave this opportunity

[sandwwraith]

I think it is OK to have it if the message itself is not malformed