scorch uses prealloc'ed buffer for docNumberToBytes()

On a couple of micro benchmarks on a dev macbook using bleve-query on
an index of 50K wikipedia docs, scorch is now faster than
upsidedown/moss on high-freq term search "text:date"...

       400 qps - upsidedown/moss
       404 qps - scorch before
       565 qps - scorch after

index/scorch/snapshot_index.go

@@ -365,11 +365,12 @@ func (i *IndexSnapshot) TermFieldReader(term []byte, field string, includeFreq,
 	return rv, nil
 }
 
-func docNumberToBytes(in uint64) []byte {
-
-	buf := new(bytes.Buffer)
-	_ = binary.Write(buf, binary.BigEndian, in)
-	return buf.Bytes()
+func docNumberToBytes(buf []byte, in uint64) []byte {
+	if len(buf) != 8 {
+		buf = make([]byte, 8)
+	}
+	binary.BigEndian.PutUint64(buf, in)
+	return buf
 }
 
 func docInternalToNumber(in index.IndexInternalID) (uint64, error) {

index/scorch/snapshot_index_doc.go

@@ -36,7 +36,7 @@ func (i *IndexSnapshotDocIDReader) Next() (index.IndexInternalID, error) {
 		next := i.iterators[i.segmentOffset].Next()
 		// make segment number into global number by adding offset
 		globalOffset := i.snapshot.offsets[i.segmentOffset]
-		return docNumberToBytes(uint64(next) + globalOffset), nil
+		return docNumberToBytes(nil, uint64(next)+globalOffset), nil
 	}
 	return nil, nil
 }

index/scorch/snapshot_index_tfr.go

@@ -49,8 +49,7 @@ func (i *IndexSnapshotTermFieldReader) Next(preAlloced *index.TermFieldDoc) (*in
 			// make segment number into global number by adding offset
 			globalOffset := i.snapshot.offsets[i.segmentOffset]
 			nnum := next.Number()
-			rv.ID = docNumberToBytes(nnum + globalOffset)
-
+			rv.ID = docNumberToBytes(rv.ID, nnum+globalOffset)
 			i.postingToTermFieldDoc(next, rv)
 
 			i.currID = rv.ID