Bugfixes:

Features:
        -try and test new libnf versions (> 1.16), check and reflect changes

Optimizations:
        -optimization for ascending sort, there is possibility to vastly reduce
         number of read/transfered records. If the key value of first records
         is X and the key value of Nth record is also X, it is unnecessary to
         read further records. Similar optimization is possible for descending
         sort.
        -OpenMP reduction for summaries (needs OpenMP 4.0, GCC 4.9)
        -minimize lnf_rec_read() calls for summaries

Tests:
        -add tests for reading multiple input files at once to test
         multithreading
        -add tests for reading single/multiple very large files to test
         buffer switching
        -add tests for reading nonexistent/invalid/empty files in all modes.
         There is a possibility of deadlock.

Documentation:
        -clarify repeating options
        -create a ChangeLog

Others:
        -create and use more MPI tags
        -suppress printing command-line error messages on all nodes
        -runtime debugging output
