Entity usage should use the processed text value and not the raw text value

Now to add test coverage...

I've been thinking about this issue and it makes me wonder if we should be maintaining:

• \Drupal\entity_usage\Plugin\EntityUsage\Track\EntityEmbed
• \Drupal\entity_usage\Plugin\EntityUsage\Track\HtmlLink
• \Drupal\entity_usage\Plugin\EntityUsage\Track\MediaEmbed
• \Drupal\entity_usage\Plugin\EntityUsage\Track\LinkIt
• \Drupal\entity_usage\Plugin\EntityUsage\Track\CkeditorImage

I think we should have a single text field plugin that uses the processed text (and summary if available) to determine usage. The plugin should support entity usage discovery from both the URL and the data-entity-uuid and data-entity-type attributes.

This approach would have the following advantages:

• No double reporting from LinkIt/ EntityEmbed / CkEditorImage / MediaEmbed - these plugins are all essentially the same - they use the attributes to determine if there is a link
• A single plugin working through all the links is more performant and less code to maintain
• Support for any esoteric filter that adds links in a way not covered by the existing plugins

@marcosano has pointed out that entity embed is a problem. Entity embed possibly will add links to other entities and we wouldn't want these usages registered as part of the entity with text field... so yeah #4 is wrong-headed.

That said I think it points to the fact that we could reorganise the plugin relationships so we only have to create the dom for each text field once.