Indexer crashes when decoding large size secondary keys during scan

The scan decode buffer size is initialized to 16KB block buffer.
If a secondary key decoded size is more than 16KB, the collatejson decode will fail.
We have to make decoded buffer to auto resize if the decode size is bigger.

scan_pipeline.go:161
sk, docid, count = siSplitEntry(row, tmp)¬