Nullness check FPs and FNs

[JuditKnoll]

There are some problems with the Null Value Analysis.
The Ideas_2011_07_22 file tested in RegressionIdeas20110722Test shows these problems.
There should be the following buginstances found:

• 4 RCN_REDUNDANT_NULLCHECK_WOULD_HAVE_BEEN_A_NPE
  • at getHashCode3() method line 34
  • at getHashCode4() method line 41
  • at getHashCode5() method line 48
  • at getHashCode6() method line 55
• 2 NP_PARAMETER_MUST_BE_NONNULL_BUT_MARKED_AS_NULLABLE
  • at getHashCode6() method parameter x
  • at getHashCode7() method parameter x
• 0 NP_NULL_ON_SOME_PATH

Instead these buginstances are found:

• 4 NP_NULL_ON_SOME_PATH
  • at getHashCode3() method line 37
  • at getHashCode4() method line 44
  • at getHashCode5() method line 51
  • at getHashCode6() method line 58

This makes formers FNs, and the latters FPs.
RCN_REDUNDANT_NULLCHECK_WOULD_HAVE_BEEN_A_NPE and NP_NULL_ON_SOME_PATH are reported by edu.umd.cs.findbugs.detect.FindNullDeref, with the latter also being reported by edu.umd.cs.findbugs.detect.FindNullDerefsInvolvingNonShortCircuitEvaluation.
NP_PARAMETER_MUST_BE_NONNULL_BUT_MARKED_AS_NULLABLE is reported by edu.umd.cs.findbugs.detect.NoteUnconditionalParamDerefs and edu.umd.cs.findbugs.detect.InconsistentAnnotations.
All of these are connected to nullness, and I think the root cause is in IsNullValueAnalysis.

Currently it doesn't break the build, because when running the tests some classes needed for the analysis are missing, which is solved by #2840. However, analysing a simple project containing Ideas_2011_07_22 only riports the NP_NULL_ON_SOME_PATH FPs.