When a page provides more specific guidance, ignore rel="canonical".

A website might want search engines to treat multiple versions of a page as the same (for example, when creating multiple versions of a page), but might want Hypothesis to differentiate the different versions.

Right now it's unclear, when using both rel="canonical" and a method used to avoid document equivalence, whether Hypothesis would treat the documents as equivalent or not. We should set a single standard for what happens when you employ multiple equivalence methods.

**Currently:**
Page A and page B are two versions of the same article. The website owner wants the same rel="canonical" address for both pages, but wants the annotations on each article to remain distinct.

In addition, rel="canonical" has caused confusion for publishers in the past. As examples:

- https://github.com/hypothesis/product-backlog/issues/263
- https://hypothesis.zendesk.com/agent/tickets/6389
- https://hypothesis.zendesk.com/agent/tickets/3297

**What to do:**
When a page uses both rel="canonical" and another method of establishing or avoiding document equivalence, ignore rel="canonical" and use the other method.

https://hypothesis.zendesk.com/agent/tickets/8531