new: 'query' option in client constructor and per request options

[DonutEspresso]

Went to try and fix #134 but could not repro. But added unit tests to make sure this is first class and supported moving fwd.

[coveralls]

Coverage increased (+0.2%) to 86.202% when pulling 74dc8e1 on query into c6fd356 on master.

[coveralls]

Coverage increased (+0.1%) to 86.166% when pulling f447c52 on query into bb65be9 on master.

[coveralls]

Coverage increased (+0.1%) to 86.166% when pulling f447c52 on query into bb65be9 on master.

[hekike]

Can you add a test that covers a situation when both ? in the url and the query object is passed. I think the documentation should also cover this case.

[hekike]

As we have the parsed url technically we could also merge the query objects. Not saying we should, just make sure the expected behavior is documented.

[DonutEspresso]

That's a good point, on client constructor we actually don't take default query options, but no reason we couldn't. If anything, it might be more consistent if we do so as it matches behavior of other options. I'll make the change.

[DonutEspresso]

Updated! Per request values override default values. While merging query params seems useful in some scenarios, I think it's easier to override and push any custom merging logic into userland.