UnicodeUtil: Fix UTF8toUTF16WithFallback byte[] overload to use fallback path, #1024

The byte[]/offset/length overload was delegating to the throwing
UTF8toUTF16 variant, so callers like BytesRef.Utf8ToStringWithFallback()
would throw DecoderFallbackException on invalid UTF-8 instead of
substituting U+FFFD as the method name and docs promise. Delegate to
the ReadOnlySpan<byte> fallback overload and add a regression test
covering the 4-arg signature with a non-zero offset.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

Author: paulirwin

src/Lucene.Net.Tests/Util/TestUnicodeUtil.cs

@@ -377,6 +377,25 @@ public void TestUTF8toUTF16WithFallback(byte[] utf8, string expected)
             Assert.AreEqual(expected, scratch.ToString());
         }
 
+        [Test]
+        [LuceneNetSpecific]
+        [TestCase(new byte[] { 0x63, 0x61, 0xc3 }, "ca\ufffd")] // ca�, start of 2-byte sequence
+        [TestCase(new byte[] { 0x63, 0x61, 0xe3 }, "ca\ufffd")] // ca�, start of 3-byte sequence
+        [TestCase(new byte[] { 0x63, 0x61, 0xf3 }, "ca\ufffd")] // ca�, start of 4-byte sequence
+        [TestCase(new byte[] { 0x63, 0x61, 0xc3, 0xb1, 0x6f, 0x6e }, "cañon")]
+        public void TestUTF8toUTF16WithFallback_ByteArrayOverload(byte[] utf8, string expected)
+        {
+            // Regression: the 4-arg byte[] overload must delegate to the fallback implementation,
+            // not the throwing one. Exercise it via a non-zero offset to confirm the arguments flow through.
+            var scratch = new CharsRef();
+            var padded = new byte[utf8.Length + 2];
+            Array.Copy(utf8, 0, padded, 2, utf8.Length);
+
+            UnicodeUtil.UTF8toUTF16WithFallback(padded, 2, utf8.Length, scratch);
+
+            Assert.AreEqual(expected, scratch.ToString());
+        }
+
         [Test]
         [LuceneNetSpecific]
         [Repeat(100)]

src/Lucene.Net/Util/UnicodeUtil.cs

@@ -1065,7 +1065,7 @@ public static void UTF8toUTF16(ReadOnlySpan<byte> utf8, CharsRef chars)
         // TODO: broken if chars.offset != 0
         public static void UTF8toUTF16WithFallback(byte[] utf8, int offset, int length, CharsRef chars)
         {
-            UTF8toUTF16(utf8.AsSpan(offset, length), chars);
+            UTF8toUTF16WithFallback(utf8.AsSpan(offset, length), chars);
         }
 
         /// <summary>