RFR 8151546: nsk/jvmti/RedefineClasses/StressRedefine fails in hs nightly

[Coleen Phillimore]

Summary: Constant pool merging is not thread safe for source_file_name.

This change includes the change for the following bug because they are 
tested together.

8148772: VM crash in nsk/jvmti/RedefineClasses/StressRedefine: assert 
failed: Corrupted constant pool
Summary: ConstantPool::resolve_constant_at_impl() isn't thread safe for 
MethodHandleInError and MethodTypeInError.

The parallel constant pool merges are mostly harmless because the old 
methods constant pool pointers aren't updated.  The only case I found 
where it isn't harmless is that we rely on finding the 
source_file_name_index from the final merged constant pool, which could 
be any of the parallel merged constant pools.  The code to attempt to 
dig out the name from redefined classes is removed.

open webrev at http://cr.openjdk.java.net/~coleenp/8151546.01/webrev
bug link https://bugs.openjdk.java.net/browse/JDK-8151546

Tested with rbt, java/lang/instrument tests, com/sun/jdi tests.  I tried 
to write a test with all the conditions of the failure but couldn't make 
it fail (so noreg-hard).

Thanks,
Coleen