AK: Use simdutf when appending UTF-16 to StringBuilder

Adds a fast path for valid UTF-16 using `simdutf`, and fall back to
the slow path for unmatched surrogates.

AK/UnicodeUtils.h

@@ -78,4 +78,50 @@ template<FallibleFunction<char> Callback>
     return -1;
 }
 
+/**
+ * Compute the maximum number of UTF-8 bytes needed to store a given UTF-16 string, accounting for unmatched UTF-16 surrogates.
+ * This function will overcount by at most 33%; 2 bytes for every valid UTF-16 codepoint between U+100000 and U+10FFFF.
+ */
+[[nodiscard]] static inline size_t maximum_utf8_length_from_utf16(ReadonlySpan<u16> code_units)
+{
+    // # UTF-8 code point -> no. UTF-8 bytes needed
+    // U+0000   - U+007F   => 1 UTF-8 bytes
+    // U+0080   - U+07FF   => 2 UTF-8 bytes
+    // U+0800   - U+FFFF   => 3 UTF-8 bytes
+    // U+010000 - U+10FFFF => 4 UTF-8 bytes
+
+    // # UTF-16 code unit -> no. UTF-8 bytes needed
+    // 0x0000 - 0x007f [U+000000 - U+00007F] = 1 UTF-8 bytes
+    // 0x0080 - 0x07ff [U+000080 - U+0007FF] = 2 UTF-8 bytes
+    // 0x0800 - 0xd7ff [U+000800 - U+00FFFF] = 3 UTF-8 bytes
+    // 0xd800 - 0xdbff [U+010000 - U+10FFFF] = 4 UTF-8 bytes to encode valid UTF-16 code units,
+    //                                         or 3 UTF-8 bytes to encode the unmatched surrogate code unit.
+    // 0xdc00 - 0xdfff [U+010000 - U+10FFFF] = 0 UTF-8 bytes to encode valid UTF-16 code units (because it is already accounted for in 0xdc00 - 0xdfff),
+    //                                         or 3 UTF-8 bytes to encode the unmatched surrogate code unit.
+    // 0xe000 - 0xffff [U+00E000 - U+00FFFF] = 3 UTF-8 bytes
+
+    // # UTF-16 code unit -> actual length added.
+    // 0x0000 - 0x007f = 1
+    // 0x0080 - 0x07ff = 2
+    // 0x0800 - 0xd7ff = 3
+    // 0xd800 - 0xdbff = 3
+    //   ^ If the next code unit is 0xdc00 - 0xdfff, they will combined sum to 6, which is greater than the 4 required.
+    //   Otherwise, 3 bytes are needed to encode U+D800 - U+DBFF.
+    // 0xdc00 - 0xdfff = 3
+    //   ^ If the previous code unit was, 0xd800 - 0xdbff, this will ensure that the combined sum is greater than 4.
+    //   Otherwise, 3 bytes are needed to encode U+DC00 - U+DFFF.
+    // 0xe000 - 0xffff = 3
+
+    size_t maximum_utf8_length = 0;
+
+    // NOTE: This loop is designed to be easy to vectorize.
+    for (auto code_unit : code_units) {
+        maximum_utf8_length += 1;
+        maximum_utf8_length += code_unit > 0x007f;
+        maximum_utf8_length += code_unit > 0x07ff;
+    }
+
+    return maximum_utf8_length;
+}
+
 }