fix: support TTC font face selection via fontconfig index

[sitiyou]

See #495 for more information.

Satty currently loads fonts through Canvas::add_font, which always uses face index 0 internally.

This PR changed the font initialization flow to:

• create a TextContext first
• load the matched font through TextContext::add_shared_font_with_index
• initialize the canvas with Canvas::new_with_text_context(...)

By doing so, Satty can use the face index returned by fontconfig and load the correct TTC face instead of always falling back to face 0

---

close #495

[sitiyou]

I have fixed the ci linter errors

[RobertMueller2]

Thank you looking into this and for the PR. I first thought this was a femtovg bug, but this change looks almost like we're not using it correctly?

I haven't tested the change fully yet, but I've got two remarks.

• can you use eprintln! instead of println! please? we're trying to correct this in fix: replace println! with eprintln! #479 for all the other places
• the two match arms do almost the same. Perhaps this could be deduplicated with a helper function e.g. with is_fallback bool parameter? What do you think?

[sitiyou]

I first thought this was a femtovg bug, but this change looks almost like we're not using it correctly?

I agree this is not really a femtovg bug.
Canvas::add_font as a simple high-level API is reasonable even without a face_index parameter, since TextContext already provides the full font-loading API.

I haven't tested the change fully yet, but I've got two remarks.

I have replaced my println! with eprintln!, and created a load_font closure that consolidates all font loading logic.

I also changed loaded_paths from HashSet<PathBuf> to HashSet<(PathBuf, u32)>, using (path, face_index) as the deduplication key so different faces in the same TTC file are not mistaken for the same font.

---

It now works great on my system. I'm happy to help with testing if you need it.

[RobertMueller2]

Makes sense and looks good to me.