fix: use byte-safe indexing in style encapsulation to prevent UTF-8 panics

[BenjaminDobler]

Summary

• Fix panic when CSS selectors contain multibyte UTF-8 characters (e.g. ü, é, ─ box-drawing chars in comments)
• Five functions in encapsulation.rs used char indices to slice &str, which panics when a char index falls inside a multibyte UTF-8 sequence
• split_by_combinators, find_pseudo_element_start, find_pseudo_class_start: switched from chars().collect() to char_indices().collect() so string slicing uses byte offsets
• try_scope_pseudo_function_with_context, find_matching_paren: switched from chars[i] to bytes[i] matching since (/) are ASCII and UTF-8 guarantees ASCII bytes never appear as continuation bytes

Reproduction

Any SCSS file with multibyte UTF-8 characters processed through style encapsulation triggers the panic:

thread '<unnamed>' panicked at crates/oxc_angular_compiler/src/styles/encapsulation.rs:2279:37:
byte index 5 is not a char boundary; it is inside '─' (bytes 3..6)

Test plan

• Added 4 new unit tests covering multibyte UTF-8 in attribute selectors, pseudo-elements, pseudo-classes, and combinator splitting
• All 60 existing shadow CSS tests pass
• Full cargo test passes
• cargo fmt --check passes
• No new clippy warnings introduced (pre-existing warnings remain)

🤖 Generated with Claude Code