fix(binding-coap): send error response if content negotiation fails

[JKRhb]

Dealing a bit more with CoAP content negotiation, I noticed in the RFC that instead of using a default Content-Format in the case of failed negotiation, an error response should must be sent. This PR takes that into account, adjusting the internal negotiateContentFormat method and the test. Since this caused the logic to become a bit more complicated, I added some documentation to hopefully clarify how the new, now RFC-compliant process works.

[ektrah]

Could the whole function could be simplified like this?

switch (accept) {
    case ContentSerdes.DEFAULT:
    case ContentSerdes.TD:
    case null:
        res.code = "2.05";
        res.setHeader("Content-Format", accept || ContentSerdes.TD);
        res.end(JSON.stringify(thing.getThingDescription()));
        break;
    default:
        res.code = "4.06";
        res.end(`Content-Format ${accept} is not supported by this resource.`);
        break;
}

[JKRhb]

I incorporated your suggestion in d1f3ae1. My initial thoughts were that this method could be a generic one for content negotiation, but maybe that is a bit too much. Having a specific method could make more sense, at least for now.

[ektrah]

I think the generic solution would probably look like this:

let serializer = SERIALIZERS[accept];
if (serializer) {
    let content = serializer.serialize(thing.getThingDescription());
    res.code = "2.05";
    res.setHeader("Content-Format", content.Type);
    res.end(content.Data);
} else {
    res.code = "4.06";
    res.end(`Content-Format ${accept} is not supported by this resource.`);
}

[JKRhb]

The latest commits now implement a generic version using the ContentSerdes class for serialization. In the future, this would also allow for different TD representations (e.g., as CBOR or XML), but it seems as if these two codecs are currently not supported/implemented.

[JKRhb]

Hmm, does somebody know why the Windows tests are suddenly failing? :/ Edit: My bad, it's actually the firestore tests Edit2: Or both actually?

[danielpeintner]

it's actually the firestore tests Edit2: Or both actually?

we do have issues with firestore since a while (see #827)..

[JKRhb]

@danielpeintner Thank you for approving :) I took the freedom of squashing the commits.

[relu91]

Minor point below

[danielpeintner]

Good to go from my side. I let @relu91 push the merge button.