Fix encoding fallback error handling and error message formatting

- Raise ArgumentError("too big fallback string") when a fallback
  function returns a value that cannot be encoded to the destination
  encoding. Previously, the invalid character was silently dropped.

- Format UTF-8 codepoints as U+XXXX in UndefinedConversionError messages
  instead of showing the raw byte dump (e.g., "U+FFFD" instead of
  "\xEF\xBF\xBD"), matching MRI behavior.

These changes fix all 12 failing specs in encode_spec.rb related to
fallback option error handling.

Author: khasinski

core/src/main/java/org/jruby/util/io/EncodingUtils.java

@@ -1456,9 +1456,11 @@ public boolean call(ThreadContext context, IRubyObject fallback, EConv ec) {
                 rep = rep.convertToString();
                 Encoding repEnc = ((RubyString) rep).getEncoding();
                 ByteList repByteList = ((RubyString) rep).getByteList();
-                ec.insertOutput(repByteList.getUnsafeBytes(), repByteList.begin(), repByteList.getRealSize(), repEnc.getName());
+                int ret = ec.insertOutput(repByteList.getUnsafeBytes(), repByteList.begin(), repByteList.getRealSize(), repEnc.getName());
 
-                // TODO: check for too-large replacement
+                if (ret == -1) {
+                    throw argumentError(context, "too big fallback string");
+                }
                 return true;
             }
             return false;
@@ -1691,18 +1693,27 @@ else if (result == EConvResult.UndefinedConversion) {
             int errBytesP = ec.lastError.getErrorBytesP();
             int errorLen = ec.lastError.getErrorBytesLength();
             final byte[] errSource = ec.lastError.getSource();
-            if (Arrays.equals(errSource, "UTF-8".getBytes())) {
-                // prepare dumped form
-            }
 
             RubyString bytes = newString(context, new ByteList(errBytes, errBytesP, errorLen - errBytesP));
-            RubyString dumped = (RubyString) bytes.dump(context);
+            String charRepresentation;
+
+            // For UTF-8 source, format as U+XXXX codepoint instead of byte dump
+            if (Arrays.equals(errSource, "UTF-8".getBytes())) {
+                int codepoint = StringSupport.preciseCodePoint(UTF8Encoding.INSTANCE, errBytes, errBytesP, errBytesP + errorLen);
+                if (codepoint >= 0) {
+                    charRepresentation = String.format("U+%04X", codepoint);
+                } else {
+                    charRepresentation = ((RubyString) bytes.dump(context)).toString();
+                }
+            } else {
+                charRepresentation = ((RubyString) bytes.dump(context)).toString();
+            }
 
             mesg = new StringBuilder();
             if (Arrays.equals(errSource, ec.source) &&  Arrays.equals(ec.lastError.getDestination(), ec.destination)) {
-                mesg.append(dumped).append(" from ").append( new String(errSource) ).append(" to ").append( new String(ec.lastError.getDestination()) );
+                mesg.append(charRepresentation).append(" from ").append( new String(errSource) ).append(" to ").append( new String(ec.lastError.getDestination()) );
             } else {
-                mesg.append(dumped).append(" to ").append( new String(ec.lastError.getDestination()) ).append(" in conversion from ").append( new String(ec.source) );
+                mesg.append(charRepresentation).append(" to ").append( new String(ec.lastError.getDestination()) ).append(" in conversion from ").append( new String(ec.source) );
                 for (int i = 0; i < ec.numTranscoders; i++) {
                     mesg.append(" to ").append( new String(ec.elements[i].transcoding.transcoder.getDestination()) );
                 }