Set XML Accept header when fetching tabular data

[jwalgran]

The default initialization for instances of this class sets the Accept header to application/json. The transform_tabular_data utility function that is used to parse the response expects XML content.

Connects #62

PR #64 includes a similar fix as part of a larger change set. If #64 is merged this PR can be closed.

[anttipalsola]

@jwalgran Is there some reason to use XML in the first place? As the BambooHR API is also capable to return tabular data as JSON, why not just skip utils.transform_tabular_data altogether and just return r.json()?

[jwalgran]

The PR as written is the smallest change I could come up with that prevented the function from raising an exception. I tend to agree that since the data is being transformed and returned as a dictionary, requesting the tabular data as JSON from the start makes sense. I am not sure if the data returned from the API has the same "shape" as the dictionary currently returned by transform_tabular_data. If it does not, we would presumably want to repurpose transform_tabular_data and reshape the JSON so that the values returned from get_tabular_data remain consistent.

[anttipalsola]

The "shape" of the JSON data roughly as follows:
[[{User 1 row 1}, {User 1 row 2}], [{User 2 row 1}], ..., [{User N row 1}, ..., {User N row N}]]

So, it is not identical. I agree that we should keep the API stable, so I did what you suggested and created yet another fix for the same problem: #67. I did not modify transform_tabular_data because it is used also by other functions but created a new one (transform_json_tabular_data).

So, now we have three PRs for the same problem. I hope one of them gets merged soon. :-)