versions with no actual diffs mess with history

[petiepooo]

steps to reproduce:

1. Create a new article with reference content
2. update once with a content change and a change comment (for reference)
3. update article again, but with no change in content, only a change comment.
4. view history (only see versions 1 and 2, no version 3)
5. view article (version 3 exists only in kb_articles table, not in kb_article_versions)
6. update again with content and change comment
7. view history (only see versions 1, 2, and 4)
8. kb_article_versions table does not have version 3 or its change comment

Users should be creating comments, not entering version change comments, if they are not changing the article's content, but this still doesn't seem like correct behavior. Should the update be rejected if there is no change to the article's content?