BibRank: citesummary optimizations

• Use the key parameter for sort instead of the cmp parameter to sort by citations faster in the citesummary
• Still in the citesummary, instead of appending to lists, uses list comprehensions
• Instead of storing the whole citation dict in memory, we now only store citations counts in memory.
• Stores citation dict in the database and removes it from memory.
• Improves code quality in bibrank_citation_indexer.py.
• In citesummary, sort the citations count first, so that we only have to do it once.
• Use 2 different methods for sorting citations counts in citesummary:
  • for less than 20000 records, we use the counts dictionary and access by key
  • for more than 20000 records, we use the presorted citations counts list (closes #1481)
• Split stats computation and html rendering in the citesummary in 2 different functions.
• With that change, we can now iterate only once through, the citations counts and compute all the stats. (closes #1217)
• When citesummary was computed for less than 20000 records, we would display the number of self-citations instead of the number of citations.

Signed-off-by: Alessio Deiana <alessio.deiana@cern.ch>