A command to completely disable IPv6 RA on interface (PR)

[Sokolmish]

Hello.

Having encountered a problem described in #7738, I've found that there's no possibility to completely disable sending RA messages when extended-nexthop BGP capability is enabled (and in some other cases). I've described the way to reproduce the problem in the issue. There's a similar discussion #15994 where it's said that RA are used for unnumbered BGP peers, but we don't use them and still need the extended-nexthop capability.

After looking into the source, I came to the conclusion that the suppress-ra command is intended to enable RA (in "no" form), but not to disable it. I suggest adding a new command disable-ra that will disable it completely, disregarding other conditions, so as not to break the routing in the network.

I've prepared a patch available there (updated link), so I'm ready to make a PR with the described functionality. But since it affects public configs, I decided to discuss it here beforehand.

P.S. I've also noticed that while RA can be enabled by bgpd and vrrpd (aside from vty), there are pieces of code (e.g. BGP_RA_CONFIGURED) that assume only BGP in their names. Hypothetically, it might lead to a situation where, say, vrrpd will disable RA despite bgpd still wanting it. I haven't tried to reproduce it. Should I make an issue for it?