RFR (S) 8218755: [REDO] Symbol leak in prepend_host_package_name

[Kim Barrett]

> On Feb 12, 2019, at 10:13 AM, coleen.phillimore at oracle.com wrote:
> 
> Summary: fix Symbol refcounting again, add comment and a test.
> 
> open webrev at http://cr.openjdk.java.net/~coleenp/2019/8218755.01/webrev
> bug link https://bugs.openjdk.java.net/browse/JDK-8218755
> 
> I hope this is better.  Ran hs-tier1-6 and a new test!
> 
> Thanks,
> Coleen

So this isn't really a "REDO", but rather a fix on top of the earlier
"fix", with no "BACKOUT" of that earlier fix.  Right?

Given that the recently added update_class_name now seems to have only
one caller, is it really worth keeping?

Other than that, looks good.  Thanks for adding a test.