Preserve lone surrogate escapes in string literal emit

[Copilot]

Lone surrogate escapes in string literals were decoded to U+FFFD, corrupting emitted enum member names and const enum substitutions.

• Scanner

  • Preserve lone \uD800/\uDC00 escapes as surrogate sentinels instead of replacement characters.
  • Apply the same handling to braced Unicode escapes such as \u{D800}.

• Printer

  • Decode surrogate sentinels when escaping string literal text.
  • Emit lone surrogates as UTF-16 escape sequences.

• Regression coverage

  • Added compiler coverage for enum member names and const enum values:

    enum E {
      "\uD800lone" = 1,
    }
    
    const enum S {
      C = "\uD800",
    }
    
    const c = S.C;

  • Updated existing surrogate baselines to reflect preserved lone surrogate escapes.

[jakebailey]

#4181