Accept header handling is still incorrect

[zmeggyesi]

Following in the footsteps of #673, and referring to #487 and #488, I think the Accept header's handling should be refined further to return a 406 Not Acceptable response, instead of falling back to text/html, as dictated by RFC7231.
I realise, however, that this is a breaking change, even if made strictly opt-in via a configuration parameter.

What are the thoughts in the community on this? Would it be worth sinking some time into it, or is the current algorithm "good enough"?

[raphaelbauer]

I have no particular opinion :) For the services I am creating I never had to deal with super fine-grained accept headers...