Fixes according to feedback

Use email from keys.txt

• Check that gpg is installed
• Unset unencrypted Conduit token after use

Update same base image used for CI builds rather than introducing a new one.

In D6291#155387, @deadalnix wrote:

I would rather avoid taking ownership of this. The part of libsecp256k1 we have ownership over are very well delimited and separated from the rest of the codebase. IMO this is a feature.

I would suggest you go about this by either:

• Disable this warning for libsecp256k1 and report the problem upstream.
• Submit a patch upstream.

In D6265#154390, @deadalnix wrote:

Why would we want the land bot to use different dependencies that CI builds?

Are the missed changes in LocaleDependenceLinter due to missing backports? If so, why not do those first?

In D6298#155315, @nakihito wrote:

Should probably start including checking the seeder version?

Fixed output ordering issue + tests

Fix nits

Rebase

Talked offline to clear up when splitting up backports is reasonable. Since there's a large stack we'll let it slide for now.

Any reason to not do the whole backport?

Thanks for cleaning this up :)

My original feedback still hasn't been entirely incorporated. Look at the name (-connect) and the description you've given it (essentially: "do some things related to seed lists but say absolutely nothing about connecting"). This is incredibly confusing for a user that has no idea how a seeder works. They know nothing about any default seeds.

abc-invalid-message will eventually be replaced by p2p_invalid_messages once we've backported all of the test cases. But as of this patch, it still provides test coverage that p2p_invalid_messages does not.