Multiaddress validation by IR

[carpawell]

In #549 there was a chain of actions: TLS was added => changed the argument of client cache's Get() method from string to network.Address to define what exactly connection should be created => changed client cache's key's format from "HostString" to multiaddr => "/dns4/s01.neofs.devenv/tcp/8080/tls" and "/tls/dns4/s01.neofs.devenv/tcp/8080" are logically the same keys(and both are correct for go-multiaddr library), but there will be two clients created for them

Therefore, it should be discussed, if IR should validate(to be in the same format) multiaddress on bootstrap or not.

[alexvanin]

Another option is to define canonical multiaddress field order and convert all address to it locally in client cache.

[fyrchik]

I think the former is correct and we can (should) disallow the latter. I haven't found an explicit description in a doc, but
Encapsulate method appends wrapped protocol to the end https://github.com/multiformats/go-multiaddr/blob/master/multiaddr.go#L140
Here is example in doc https://github.com/multiformats/go-multiaddr/blob/master/doc.go#L29

Second example (/tls/dns4/...) means something like "ip4 over TLS" which is certainly not what we want. It is not difficult to validate address on a client especially when most of the parsing will be hidden in a SDK.

[alexvanin]

• Define expected format of multi address in neofs-spec (@realloc will find best place where to put it).
• Implement sanity check of multiaddress according to this format.