plperl fix issues

I am working on the changes necesssary to fix the plperl issue raised a
week or so ago. I have code that can detect if we can safely run two
interpreters. However, things are complicated by the fact that plperl.c
uses a couple of perl hashes to store data: plperl_proc_hash and
plperl_query_hash. The complication arises because we would need to keep
switching perl contexts in order to access them. Rather than add this
complication I am tempted to replace them with standard postgres
dynahashes. Can anyone see any objection?

cheers

andrew