Drop Support for Preserving the 'Payload's of Unknown Variants

[InsertCreativityHere]

For variant enums marked unchecked, we allow decoding the values of "unknown enumerators",
and when we do, we stuff it's data into a special Unknown enumerator which looks something like this:

public partial record class Unknown
{
    int Discriminant,
    global::System.ReadOnlyMemory<byte> Fields,
}

I propose we remove the 'Fields' property from it.

1. We don't preserve the bytes from tags that we skip, and we recognized all of our old preserve-slicing logic as not useful. I don't see why the bytes of unknown enumerators is different.
2. Either:

• The receiver has no idea what this discriminant is => these bytes are useful to them
• The receiver knows this discriminant => the correct fix is to add this enumerator to your Slice definitions

There is no case where these bytes can be useful, except for forwarding them along. But I find this very niche.

---

It is easy to add this preservation back in the future in case we reverse course, or someone wants it.

[pepone]

A downside of this proposal is that if you sent the unknown variant to a peer that know about the variant it would result in a decoding exception, because you sent a discriminant without its expected fields. It makes more sense to keep things as is.

[externl]

It's not much different than how we drop unknown tags when decoding structs. You lose those tags when re-encoding the struct.

[InsertCreativityHere]

This idea of dropping the encoded fields is indeed, not worth implementing. If we want to preserve unknown variants (with a special Unknown variant), then we might as well also preserve it's discriminant and it's payload.

---

It's not much different than how we drop unknown tags when decoding structs. You lose those tags when re-encoding the struct.

This is really a point towards dropping support for unknown-variant-preservation completely. Not just removing the payload from it. Either we fully preserve the unknown variant (so it can be re-encoded), or we shouldn't preserve it at all. And treat it like tags: if we get an unknown variant, we don't try to preserve it. That would be my preference! But slimming down individual fields buys us nothing.