Ensure we never use (0,0) SourceSections

* Those would appear in the debugger as going back to the first line,
  which is incorrect and confusing.

Author: eregon

src/main/java/org/truffleruby/language/RubyNode.java

@@ -106,6 +106,10 @@ public void unsafeSetSourceSection(SourceSection sourceSection) {
     public void unsafeSetSourceSection(int charIndex, int sourceLength) {
         assert !hasSource();
 
+        // The only valid case for a (0,0) SourceSection is the RootNode SourceSection of eval("").
+        // We handle that case specially by using an unavailable SourceSection, so then every (0,0) is a bug.
+        assert !(sourceLength == 0 && charIndex == 0);
+
         setSourceCharIndex(charIndex);
         setSourceLength(sourceLength);
     }

src/main/java/org/truffleruby/parser/YARPTranslatorDriver.java

@@ -177,7 +177,9 @@ public RootCallTarget parse(RubySource rubySource, ParserContext parserContext,
 
         var node = parseResult.value;
 
-        final SourceSection sourceSection = source.createSection(0, rubySource.getBytes().length);
+        final SourceSection sourceSection = rubySource.getBytes().length == 0
+                ? source.createUnavailableSection()
+                : source.createSection(0, rubySource.getBytes().length);
         final SourceIndexLength sourceIndexLength = SourceIndexLength.fromSourceSection(sourceSection);
 
         final String modulePath = staticLexicalScope == null || staticLexicalScope == context.getRootLexicalScope()