gc_massacre branch is ready for testing.

[Vasily Chekalkin]

Hello.

Old GC behave itself in really weird way passing this test. Test fixed
in r49217 in branch.

FileHandle PMC is actually alive because it's referenced by
CallContext from $P0.'open'() method call. So, it shouldn't be
destroyed during mark&sweep. Old GC doesn't do full trace and missed
that object is actually alive and destroy it.

-- 
Bacek.

>> $ prove -v t/pmc/filehandle.t
>> t/pmc/filehandle.t ..
>> 1..23
>> ok 1 - new
>> ...
>> ok 21 - exit status
>> not ok 22 - timely destruction
>>
>> # Failed test 'timely destruction'
>> # at t/pmc/filehandle.t line 739.
>> # got: ''
>> # expected: 'a line
>> # '
>> ok 23 - get_bool
>> # Looks like you failed 1 test of 23.
>> Dubious, test returned 1 (wstat 256, 0x100)
>> Failed 1/23 subtests
>> (less 1 skipped subtest: 21 okay)