Improve parsing of HTML anchor content

[sorcio]

This attempts to address #53.

1. Fixes case 1 by skipping child elements of any link anchor
2. Fixes case 2 by normalizing whitespace (at least in the easy case) of anchor textContent

I'm not sure if this has any significant performance impact that should be evaluated. I tried to paste some large content and at least I didn't notice it being any slower than before. Calling a node filter on every element in the tree could be expensive, but maybe not too much?

This does not address the <br> case. I guess the only way to deal with it would be to traverse the children of the anchor and collect the text, but even that would probably not cover all the cases. And I suspect it might be too expensive if somehow it happens that the tree under the anchor element is very complex.

I also realized that using innerText instead of textContent doesn't work here (while normally it makes a difference), because when the element is not being rendered the DOM doesn't care about the semantics of <br>.

Please just take this as a suggestion. I mostly don't understand what I'm doing but wanted to give this a try anyway.

[manuelpuyol]

Thanks for working on it @sorcio !