After the discussions in #44 I have decided to look into it, and ended up refactoring the whole loader.
I did that with 3 goals:

1. make the code easier to understand
   Therefore, I've split loader.js into 2 new sub-loaders to make it more obvious what is happening
   I've also renamed some variables and added a few comments
2. add a <template> tag
   turned out to be a very simple 1 line change
3. allow multiple tags (eg 2 <script> tags)
   Tags now don't really matter much anymore. The only rule now is to have at least 1 script tag which will be the primary javascript (or typescript, coffee, ...) part (either the tag with id="component" or the first script tag). All other parts will be loaded into that via require() regardless of tag.
   This now means that you could in theory even have <script type="text/css">body { background: red; color: blue; }</script> and it would load correctly via the css-loader & style-loader because only the type attribute is used to decide what to do. It's silly, but it would work. Primary benefit of the tags now is just the different syntax highlighting in editors.

Because it made sense in my refactoring, I also changed the return value of the parser, and thereby fixed #52

EDIT: <template> is actually not working as intended at the moment. We need to turn the parsed tree back into a string in order to work.

This PR is not ready to be merged, but rather a request for review and comments

TODOs:

• update tests
• add more documentation
• fix <template>: turn tree object into string
• verify that this actually satisfies the usecase of "jest" unit tests