csmount._chunk_read()

• read fragments in parallel
• need to verify which ones are "alive"
  • perhaps chunkserver.py module should verify 'alive'
• need to re-attempt fragment read if we have a "mid-read" failure