cli: add --http-addr flag

[glycerine]

For security, the --http-addr flags makes the Admin UI
bind an IP address that can be private or distinct from
the cluster (such as 127.0.0.1).

Fixes #7474.

---

This change is

[glycerine]

I don't grok the test failures from CI.
Update: oh I think I get it. The docker container acceptance tests depend on the UI being on the expected IP that is the same as the host/container. I kept the default the same to fix this.

[knz]

@tamird @mberhault what's the best place to add a test for this flag in our acceptance tests?

[mberhault]

small nit: use HTTPAddrName, it doesn't have to be an IP address, dns names work fine.

[mberhault]

Actually, a bigger nit: let's stick to the same naming scheme as {host, port}, where we now have {http-host, http-port}.

[tamird]

Please hold off on merging this - it's going to conflict with multiple addresses / external address work, and the semantics aren't completely clear (i.e. we may want to move to specifying full addresses rather than addresses and ports).

[mberhault]

A test would be nice, and the acceptance tests are probably the best place since we would ideally like some networking isolation.

@tamird: thoughts on the tests? Also how this may fit in with future post modifications?

[glycerine]

Marc - I agree a test or two would be nice. I had the worst time trying to get the acceptance suite to run locally. I filed #7477 about that.

Tamir - okay, let me know when I might be of assistance.

[glycerine]

@tamird Have the semantics been decided? I'd like to move forward on this.

Working at adding a test and fixing the nits above...

[glycerine]

I've added a surgically precise test/unit test. I think an acceptance test is beyond the scope of what I, with my limited resources, can attempt at this point.

Also, I've addressed the nits.

@tamird, @mberhault, @knz, would you be okay with proceeding? It's a very small change, and it fixes a very big security hole. This would unblock me from serious further testing.

If there's a big change later towards combined IPs and ports, this really doesn't add much more to change.

[knz]

---

Reviewed 3 of 3 files at r3.
Review status: all files reviewed at latest revision, 2 unresolved discussions, all commit checks successful.

---

Comments from Reviewable

[knz]

@glycerine I'll take that one. The fact we want to do things perhaps in a different way in the future does not prevent us from accepting your patch now, for use in the interim.

[glycerine]

Thanks @knz.

[glycerine]

@jseldess Now that cockroachdb/docs#443 merged, I think the docs-todo label can be removed.

[jseldess]

Thanks, @glycerine. Just tweaking a few minor things before publishing (probably early tomorrow).