This patch removes a FIXME regarding the retrieval
of the disk template for computation of the cluster
information for the iallocator. After examining the
code, I figured out that it is actually a quite
robust way to just check if the disk template is
available, in particular with respect to extensions
of the iallocator.

In this context, I also updated the man pages of
the operations that use the iallocator, because so
far it was not well documented which disk template
is considered when more than one instance was
allocated. The situation is not ideal (the default
disk template is used), but it is still better than
in 2.9. where only LVM storage was considered.
With the implementation of storage pools, it would
make sense to extend the iallocator protocol to
include storage information for more than one
template.

Signed-off-by: Helga Velroyen <helgav@google.com>
Reviewed-by: Klaus Aehlig <aehlig@google.com>