ipv6: do not use per netns icmp sockets

Back in linux-2.6.25 (commit 98c6d1b2 "[NETNS]: Make icmpv6_sk per namespace.",
we added private per-cpu/per-netns ipv6 icmp sockets.

This adds memory and cpu costs, which do not seem needed.
Now typical servers have 256 or more cores, this adds considerable
tax to netns users.

icmp sockets are used from BH context, are not receiving packets,
and do not store any persistent state but the 'struct net' pointer.

icmpv6_xmit_lock() already makes sure to lock the chosen per-cpu
socket.

This patch has a considerable impact on the number of netns
that the worker thread in cleanup_net() can dismantle per second,
because ip6mr_sk_done() is no longer called, meaning we no longer
acquire the rtnl mutex, competing with other threads adding new netns.

Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>