Use CaseInsensitiveString for Headers

[joneshf]

Re: #136

Following the discussion in the above issue, this converts from Object String to Map CaseInsensitiveString String. Let me know if anything should be done differently.

Also addresses part of #113.

[cprussin]

CI is broken right now, I need to fix it, so please ignore the failures... sorry about that!

[cprussin]

@joneshf Sweet! This is really great, thanks a bunch!

Just one question, I don't think there's any value any more in making Headers a newtype. The only reason it was originally a newtype was so that it could have an instance of Lookup that did case-insensitive lookup, but now that Headers is a Map CaseInsensitiveString String there's no reason to have a unique instance. Thoughts?

[joneshf]

Sounds good to me.

[joneshf]

Eh, most of the tests in Test.HTTPure.HeadersSpec will break:

1) HTTPure Headers show is a string representing the headers in HTTP format
  "(fromFoldable [(Tuple (CaseInsensitiveString \"Test1\") \"1\"),(Tuple (CaseInsensitiveString \"Test2\") \"2\")])" ≠ "Test1: 1\nTest2: 2\n\n"
2) HTTPure Headers append when there is a duplicated key uses the last appended value
  (fromFoldable [(Tuple (CaseInsensitiveString "Test") "foo")]) ≠ (fromFoldable [(Tuple (CaseInsensitiveString "Test") "bar")])
3) HTTPure Headers empty is an empty Map in an empty Headers
  "(fromFoldable [])" ≠ "\n"
4) HTTPure Headers header creates a singleton Headers
  "(fromFoldable [(Tuple (CaseInsensitiveString \"X-Test\") \"test\")])" ≠ "X-Test: test\n\n"
5) HTTPure Integration runs the middleware example
  "middleware" ≠ "router"

Some of this is the different behavior of Show and others are different behavior for Semigroup. Still worth it.

[cprussin]

Ah, right. I'd forgotten about those differences. Thank you for checking it out!

[joneshf]

Still worth it.

Sorry, this was supposed to be, "Still worth it?"

[cprussin]

I think that we could clean up these APIs a bit and get rid of the newtype, but I think it's out of scope for this change. So let's go ahead and merge this and I'll add a separate ticket to remove the newtype.