This uses capybara-validate_html5 to validate HTML when parsing
response bodies in the web route tests. This found numerous
HTML validation issues:
1. <form> directly inside <tr> (many cases)
2. <p> nested inside <p> (nesting not allowed, so second <p>
implicitly closes first <p>, and second </p> is therefore
invalid) (2 cases)
3. Use of </br> instead of <br> (1 case)
4. Missing </div> (1 case)
5. Plain text responses with text/html content-type (2 cases)
To fix the first case, the forms in question are moved inside <td>.
If thtere are inputs in other <td> in the same <tr>, the <form> has
an id attribute, an the other inputs use a form attribute to tie
them to the correct <form>.