Ensure that absent bone names work

[appgurueu]

Achieved by refactoring the code to use std::optional<std::string> instead of core::stringc.

Soft prerequisite for gltf support: gltf does not require bones to be named. We want to support models with "unnamed" bones.

Corresponding Minetest changes: minetest/minetest#14330

How to test

1. Read carefully (the changes should be pretty straightforwardly fine "per construction")
2. Regression tests: Do .b3d models like Sam still work? .x models (like our "cool guy" in the demo)? Attachments to bones?

[sfan5]

Wondering why you didn't just make it a std::string? or is "" somehow reserved?

[appgurueu]

Wondering why you didn't just make it a std::string? or is "" somehow reserved?

Two reasons:

1. I'm not sure whether the empty string would be an admissible bone / scene node name. I don't see a definitive reason why it shouldn't work, at least. If it is, I don't want to mess with that.
2. std::optional forces me to handle the "absence" case everywhere. It makes it explicit when a name is absent, forcing me to handle that correctly, whereas "empty name means absence" is just a convention that's easy to (accidentally) violate.