Fix handling of indented HTML in DOMParser

[mattmikolay]

This PR fixes a bug that causes DOMParser to throw an error when parsing indented HTML with the default collapsing of whitespace.

The root cause of the bug is in add_text_node. Notice the the assignment of node_before on line 571:

prosemirror-py/prosemirror/model/from_dom.py

Line 571 in 34c9614

node_before = top.content[-1]

When top.content is an empty list, the following error is thrown:

IndexError: list index out of range

Compare with the JavaScript implementation of addTextNode in the original ProseMirror:

 let nodeBefore = top.content[top.content.length - 1]

Here, when top.content is an empty array, nodeBefore will be set to top.content[-1]. In JavaScript, a negative indexing of an empty array will correctly return undefined. However, in Python, a negative indexing of an empty list will throw an error. Thus, we need to check if top.content is truthy before accessing top.content[-1].

Thanks for the great library!

[p7g]

Thanks for the contribution!