[pyrex41]

Summary

• make pos index Unicode scalar values instead of UTF-8 bytes
• make tlstr remove one complete Unicode scalar instead of slicing at byte 1
• make string->n return the first scalar value, consistent with n->string
• add multibyte regression coverage for emoji and accented characters

Problem

(tlstr "😀é") previously sliced the UTF-8 encoding after its first byte and failed with tlstr: produced non-UTF-8 result. pos and string->n used the same byte-oriented assumption, producing inconsistent behavior for multibyte text.

Verification

• cargo test -p shen-rust --test primitives_coverage string_ops string_op_errors
• cargo fmt --all -- --check
• rebuilt release launcher:
  • (tlstr "😀é") => "é"
  • (string->n "😀") => 128512
• broader cargo test -p shen-rust run was green through all reported library and initial AOT suites; the long-running command was interrupted after its output consumer detached.