As suggested by Paolo Giarrusso, add a "likely()" to Py_DECREF, telling
GCC it doesn't call the destructor in the common path.

This optimization seems to interact well with #4896 and to be slower on
top of #4753 on my system. Patching ceval.c instead of object.h seems to
give the same result, FWIW. Benchmarks most welcome :)

Paolo Giarrusso wrote:
"Anyway, yesterday I was looking at the assembly and I realized that
GCC optimizes the case where Py_DECREF calls the destructor! Since I
don't have time for this, can you hint GCC for the better by adding
unlikely() to Py_DECREF, at least (or maybe only) for ceval.c, or
maybe even only for the main interpreter?"

pybench:
    Rounds: 10
    Warp:   10
    Timer:  time.time

    Machine Details:
       Platform ID:    Linux-2.6.24-23-generic-i686-with-debian-lenny-sid
       Processor:      Intel Celeron M 410

    Python:
       Implementation: CPython
       Executable:     /home/ajaksu/py3k/python
       Version:        3.1.0
       Compiler:       GCC 4.2.4 (Ubuntu 4.2.4-1ubuntu3)
       Bits:           32bit
       Build:          Jan 13 2009 23:40:05 (#py3k:68571M)
       Unicode:        UCS2


Test                             minimum run-time        average  run-time
                                 this    other   diff    this    other 
 diff
-------------------------------------------------------------------------------
          BuiltinFunctionCalls:   147ms   163ms   -9.7%   148ms   164ms
 -10.0%
           BuiltinMethodLookup:   138ms   154ms  -10.3%   139ms   155ms
 -10.6%
                 CompareFloats:   185ms   191ms   -3.4%   185ms   192ms
  -3.4%
         CompareFloatsIntegers:   304ms   294ms   +3.1%   304ms   296ms
  +2.7%
               CompareIntegers:   287ms   281ms   +2.0%   288ms   281ms
  +2.3%
        CompareInternedStrings:   189ms   217ms  -12.7%   192ms   217ms
 -11.7%
                  CompareLongs:   167ms   162ms   +3.3%   168ms   162ms
  +3.3%
                CompareStrings:   152ms   178ms  -14.2%   156ms   179ms
 -12.7%
    ComplexPythonFunctionCalls:   192ms   202ms   -5.1%   193ms   206ms
  -6.2%
                 ConcatStrings:   325ms   327ms   -0.7%   326ms   330ms
  -1.1%
               CreateInstances:   200ms   209ms   -4.4%   204ms   210ms
  -2.9%
            CreateNewInstances:   149ms   156ms   -4.4%   151ms   161ms
  -5.8%
       CreateStringsWithConcat:   255ms   272ms   -6.1%   258ms   275ms
  -6.3%
                  DictCreation:   145ms   151ms   -4.0%   146ms   154ms
  -5.3%
             DictWithFloatKeys:   209ms   224ms   -6.8%   209ms   224ms
  -6.7%
           DictWithIntegerKeys:   170ms   185ms   -8.0%   171ms   187ms
  -8.3%
            DictWithStringKeys:   155ms   172ms  -10.1%   156ms   173ms
 -10.2%
                      ForLoops:   162ms   183ms  -11.7%   162ms   183ms
 -11.4%
                    IfThenElse:   189ms   217ms  -12.7%   190ms   219ms
 -13.1%
                   ListSlicing:   156ms   156ms   -0.3%   156ms   158ms
  -1.0%
                NestedForLoops:   199ms   219ms   -9.1%   201ms   220ms
  -8.6%
      NestedListComprehensions:   222ms   245ms   -9.5%   228ms   251ms
  -9.2%
          NormalClassAttribute:   352ms   376ms   -6.2%   353ms   378ms
  -6.6%
       NormalInstanceAttribute:   188ms   212ms  -11.1%   191ms   215ms
 -11.3%
           PythonFunctionCalls:   173ms   192ms   -9.6%   174ms   193ms
  -9.7%
             PythonMethodCalls:   213ms   226ms   -5.6%   214ms   227ms
  -5.8%
                     Recursion:   275ms   295ms   -7.0%   276ms   296ms
  -7.0%
                  SecondImport:   158ms   158ms   +0.2%   160ms   160ms
  +0.1%
           SecondPackageImport:   164ms   168ms   -2.2%   166ms   170ms
  -2.8%
         SecondSubmoduleImport:   221ms   226ms   -2.4%   223ms   229ms
  -2.9%
       SimpleComplexArithmetic:   143ms   166ms  -13.5%   144ms   166ms
 -13.4%
        SimpleDictManipulation:   269ms   296ms   -9.2%   270ms   297ms
  -9.0%
         SimpleFloatArithmetic:   150ms   177ms  -15.3%   151ms   196ms
 -23.2%
      SimpleIntFloatArithmetic:   200ms   226ms  -11.4%   203ms   232ms
 -12.4%
       SimpleIntegerArithmetic:   200ms   226ms  -11.4%   202ms   235ms
 -13.9%
      SimpleListComprehensions:   186ms   211ms  -12.0%   193ms   218ms
 -11.5%
        SimpleListManipulation:   161ms   180ms  -10.6%   161ms   181ms
 -10.9%
          SimpleLongArithmetic:   137ms   159ms  -14.0%   137ms   162ms
 -15.2%
                    SmallLists:   197ms   209ms   -5.6%   201ms   210ms
  -4.2%
                   SmallTuples:   215ms   234ms   -8.2%   216ms   236ms
  -8.2%
         SpecialClassAttribute:   544ms   566ms   -4.0%   549ms   571ms
  -3.7%
      SpecialInstanceAttribute:   188ms   211ms  -11.0%   190ms   214ms
 -11.0%
                StringMappings:   432ms   439ms   -1.6%   433ms   441ms
  -1.9%
              StringPredicates:   183ms   198ms   -7.8%   183ms   199ms
  -7.9%
                 StringSlicing:   304ms   317ms   -4.2%   310ms   321ms
  -3.3%
                     TryExcept:   138ms   154ms  -10.5%   139ms   155ms
 -10.4%
                    TryFinally:   149ms   156ms   -4.8%   151ms   159ms
  -5.1%
                TryRaiseExcept:    76ms    78ms   -2.3%    77ms    82ms
  -6.3%
                  TupleSlicing:   254ms   254ms   -0.2%   254ms   257ms
  -1.0%
                   WithFinally:   204ms   207ms   -1.1%   206ms   209ms
  -1.1%
               WithRaiseExcept:   187ms   190ms   -2.0%   188ms   192ms
  -2.2%
-------------------------------------------------------------------------------
Totals:                         10557ms 11266ms   -6.3% 10647ms 11397ms
  -6.6%

(this=likely_object.pybench, other=classic_opcode.pybench)