[css-speech-1] new value `auto` for `speak-as`

[Zhang-Junzhi]

Currently, all values of speak-as leave it up to developers to decide entirely how content should be pronounced. Now, in the era of AI, user agents should be given the opportunity to intelligently determine pronunciation.

It would be particularly beneficial for user-generated content.

Therefore, I propose adding an auto value.

[bkardell]

normal (the initial value) already allows for 'language-dependent pronunciation rules for rendering the element’s content + generally a lot is left to the underlying voice subsystems (same with Web Speech). If you used some systems this would surely use AI or some other rules today. In Web Speech, for example, the results aren't guaranteed, but most systems will say these things "correctly" as we'd expect

1. Pi is about 3.14
2. We loaded the 4x4
3. Please meet me at 3.14pm EST
4. My birthday is 2/17/1974

[fantasai]

Exactly as Brian says, the normal value already allows for this.