fix nil-reference issue in QuotedString module

[jfredett]

fixes #201, #quote will no longer throw a nil-ref when passed a nil argument.

also adds specs for Webmachine::QuotedString#quote and #unquote. I didn't see any existing specs in the spec folder. There was one case where I encoded the current behavior, but think that maybe it should not behave this way:

context 'when the string is nil' do
  subject { unquote(nil) }

  it { is_expected.to eq(nil) }
end

IMO, either this should return "", or quote(nil) should return nil, rather than "". The former is perhaps more convenient, the latter is so that the invariants

quote(unquote(str)) == str

and

unquote(quote(str)) == str

hold. Right now, if str is nil, it would get 'promoted' to an empty string. Not sure what is preferable there.

[ghost]

Right now, if str is nil, it would get 'promoted' to an empty string.

On first thought, I'd prefer unquote(nil) == nil and quote(nil) == nil, not neccessarily for backward-compatibility, but rather for avoiding possible side-effects.

OTOH, an empty header should be considered not-set by Webmachine, anyway. Any other opinions?

[seancribbs]

Strictly speaking, nil and an empty String are different things. nil should always mean the header was not present, whereas an empty string is simply an empty header. So I'd prefer @lgierth's way... quoting/unquoting nil should return nil.

The thing is, that code should only called from places where we know the argument to be a String. If it's not, the problem should be fixed in the caller.

[jfredett]

@seancribbs While working on my 'real' issue -- I seem to be getting a 415 regardless of how I have this resource set up -- it turns out that this wasn't the root-cause of my issue (the code was being executed only due to my debugging code, not webmachine itself (I think)). So I'm not sure what the call-site was/is that caused this. Regardless of whether the actual code is/should be changed, though, I think the tests are valuable and am happy to update them to encode the correct behavior if you want to make a call on that. Otherwise I'm happy to close this as it's not going to actually fix my issue.