Fix AMPFLAG and count of good variants in multi-allelic case

[PolinaBevad]

Additional test case for set AMPFLAG can be added (was not found on public data).

Description

This PR fixes issue #94 and possible issue #147 (can be tested because there are no data from issue owner). The wrong AMPFLAG was set in the multi-allelic case because it also triggered when variants on the one amplicon in one position were different (not between different amplicons).

• Changed logic of AMPFLAG. Now it will be true only if variants differ between different amplicons (implemented the comparison between lists of variants on each amplicon).
• Changed logic of counting the good variants. It must count the presence of variant on amplicons (i.e. the number of amplicon where it appears). But in multi-allelic case, we must take not the size of the list with all variants on position, but only the presence of the current variant.
• To avoid duplication of variants in output between amplicons (when identical variants occur on different amplicons), added the check for the already added variants.
• Test case for amplicon was updated.

[PolinaBevad]

Build failed on oraclejdk10 because of Travis-CI problem appeared yesterday (here: travis-ci/travis-ci#10249).
Other builds passed checks.