Retract #21096, mostly: I had misexplained the situation

Retract #21096, mostly: I had misexplained the situation
to Scott A. Crosby.  Seeing the seed value while not good
for the ultimate paranoia viewpoint is not that bad, as long
as the users are fully aware of the dangers of disclosing
the hash seed.	So hash_seed() is okay.

Being able to see the hash values (as in Java) would
be another option, but dubious: it's not that per-key
hash values themselves are bad to allow scripts to see,
but rather that hash values are just as sensitive (from
the DoSing viewpoint) as the hash seed itself (and there
usually more hash values than the one hash seed....)

p4raw-id: //depot/perl@21112