Make robust against network problems

[iimog]

While I was trying to reproduce #15 I repeatedly got the following error:

WebEnv value not found in post output
Db value not found in fetch input
[01-16 01:09:46] [ReferenceDbCreator] 
[01-16 01:09:46] [ReferenceDbCreator] ERROR: Downloading fasta sequences for batch: 3901 - 4000 failed

This occured at different batches so I assume it is a network related issue. However, a single network error should not stop the whole pipeline from working. Instead it would be desirable to have some mechanism that retries failed downloads after a short period of time (for a limited number of times). This would make the whole process much more robust against common network and connectivity issues (including NCBI rate limits).