New inference rule for `"accessMode": "textual"`

[nicolascyr]

accessMode textual is inferred, but only when accessModeSufficient textual is inferred. However, accessMode textual should be inferred when an EPUB file contains (any) content in textual format (in the same way as visual is inferred when the EPUB file contains an image or a video).

This might not be a big deal because a vast majority of EPUB files should get accessMode textual property, but I feel it's wrong to infer it only when accessModeSufficient textual is inferred (why not in all other cases?).

Example 1 :
An EPUB file without accessMode nor accessModeSufficient property, that contains text, a cover and other images
inference = accessMode visual only
This file contains text, so accessMode textual should be inferred.

Example 2 :
An EPUB file without accessMode nor accessModeSufficient property, that contains text and (only) a cover image identified as such
inference = accessMode textual, visual and accessModeSufficient ["textual"]
This file contains text, and accessMode textual is inferred but only because/thanks to the fact accessModeSufficient textual is inferred.

[HadrienGardeur]

Up until now, we've strictly inferred additional metadata based on:

• the content of the OPF
• and of the Navigation Document and/or NCX

Unlike images in <resources>, the mere presence of XHTML resources is not enough to infer "accessMode": "textual". That's why we don't infer it yet.

A good example of that would be an FXL publication where the XHTML strictly contain images with zero alt text. In this case, textOnVisual rather than textual would be the correct value.

WIth our upcoming work to add new inferences, we'll need to parse both XHTML and CSS files, which will put us in a position to infer textual for accessMode.

[HadrienGardeur]

Our next update should be capable of inferring that information. This will be based on new capabilities from the Go toolkit used by the CLI to extract Guided Navigation Documents from HTML resources.