fix: improved serializer selection

[karol-maciaszek]

Closes #589

Checklist

• Tests have been added (for bug fixes / features)
• Docs have been added / updated (for bug fixes / features)

What kind of change does this PR introduce?

Bugfix

What is the current behavior? What is the new behavior?

Current behavior:

Accept header negotiation is done in two places: in negotiator and in http-server.

• Negotiatior is selecting matching content based on Accept header and available content types.
• http-server serializes content based on Accept header and available serializers.

To reproduce:

1. spec file has content for application/json
2. request has Accept: application/xml, application/json

Results in:
Negotiator selects application/json as it matches available contents.
http-server serializes content as xml because Accept header has xml and xml serializer exists.

New behavior:

http-server serialzes the content according to what negotiator negotiated.

Does this PR introduce a breaking change?

No

[lag-of-death]

hi @StefanDywersant, please take a look at #604 (review)

[lag-of-death]

@StefanDywersant, I'll take a look when the conflict is resolved 👍

[XVincentX]

Some minor comments but it looks good.

• The fixtures should not be modified to return the charset because the OAS document does not specify it. If you apply my suggestions, it works correctly; I guess the default serialiser is still being kicked off for some reason.

• It would be better to write this test as an E2E test; you can find more informations in the README of the test-harness directory.

[lag-of-death]

@XVincentX, I understand your reasoning in #620 (comment)

Personally, I would only use the imperative style if something is not possible to do in FP. We don't really need the mayhem of mutations and exceptions being thrown throughout the code. I like your distinction here, it's like Elm/Haskell purity of the code vs side effects taken care of by the compilers, but would try to make that imperative world as small as possible still.