Lots of refactoring

[gnott]

I was tracking this issue here https://github.com/elifesciences/elife-crossref-feed/issues/139.

I decided it was time to lint and refactor before adding new features for this library to also generate Crossref peer review deposits, which might reuse some of the existing code.

My strategy here was basically:

• refactor the generate.py file to not use excessive self class attributes
• also refactor to move any functions out of the class definition that had no-self-use
• I continued to refactor some where some values could be passed in and not require the object self
• The linters were suggesting the file was too long anyway, so I started to group similar functions together into new separate modules. So I split the functions in generate.py out into 17 new separate files
• I only changed the test scenarios where I had moved a function to a new module, or some linting of whitespace and things

Please notice when reviewing this that all tests are passing, code coverage is the same (except for a reduction in lines of code drops coverage by 0.001%) and that (hopefully) in general it looks better.

There is probably more that can be done with refactoring, but this was a larger job than I intended and want to work on building a new feature before revisiting another round of linting or refactoring.

[coveralls]

Coverage increased (+0.1%) to 99.839% when pulling 8091cdb83e8f3feda3b81f100aeb197c5072b3f6 on refactor into dc9dd335a8d6a3ad58d65c126b884bdca1abed3e on develop.

[gnott]

I hope this refactored code is an improvement over the original. As I described in the PR description, it is shown to function (almost) identically according to the automated test scenarios.

With so many lines of code edited, it would be very time consuming to provide a full review of the code prior to merging.

In order for further development of this library to continue, and for this branch not to go stale, I am going to merge it in now, and I can set about to adding new features.