Fix bad total snippets counts.

Turns out we were mixing up the counts between multiple documents by
keying according to the leaf-reader-relativ docId instead of the
index-absolute docId. This led to differing total counts between
different runs of one and the same query, which is how this was
discovered. Total counts should be accurate now.

src/main/java/com/github/dbmdz/solrocr/lucene/OcrFieldHighlighter.java

@@ -83,16 +83,11 @@ public OcrSnippet[] highlightFieldForDoc(
     }
 
     Passage[] passages;
-    try (OffsetsEnum offsetsEnums = fieldOffsetStrategy.getOffsetsEnum(reader, readerDocId, null)) {
+    try (OffsetsEnum offsetsEnums =
+        fieldOffsetStrategy.getOffsetsEnum(reader, readerDocId, null)) {
       passages =
           highlightOffsetsEnums(
-              offsetsEnums,
-              indexDocId,
-              breakLocator,
-              formatter,
-              pageId,
-              snippetLimit,
-              scorePassages);
+              offsetsEnums, indexDocId, breakLocator, formatter, pageId, snippetLimit, scorePassages);
     }
 
     // Format the resulting Passages.