Add cockroach-admin dropshot server

[jgallagher]

The goal here is to use this server to run cockroach node decommission for expunged cockroach zones. For this initial PR, the only endpoint provided wraps cockroach node status; wrapping decommission is not as trivial so I figured it should go into a separate PR.

There are currently no callers of this service, but stood up an a4x2 and confirmed I could talk to it from the switch zone via curl:

root@oxz_switch:~# curl http://[fd00:1122:3344:103::3]:32222/node/status
{"all_nodes":[{"node_id":"1","address":"[fd00:1122:3344:103::3]:32221","sql_address":"[fd00:1122:3344:103::3]:32221","build":"v22.1.9-dirty","started_at":"2024-05-24T16:47:33.137256Z","updated_at":"2024-05-24T19:01:11.345263Z","locality":"","is_available":true,"is_live":true},{"node_id":"2","address":"[fd00:1122:3344:102::3]:32221","sql_address":"[fd00:1122:3344:102::3]:32221","build":"v22.1.9-dirty","started_at":"2024-05-24T16:41:23.877326Z","updated_at":"2024-05-24T19:01:10.946872Z","locality":"","is_available":true,"is_live":true},{"node_id":"3","address":"[fd00:1122:3344:102::4]:32221","sql_address":"[fd00:1122:3344:102::4]:32221","build":"v22.1.9-dirty","started_at":"2024-05-24T16:41:24.020025Z","updated_at":"2024-05-24T19:01:11.112721Z","locality":"","is_available":true,"is_live":true},{"node_id":"4","address":"[fd00:1122:3344:101::4]:32221","sql_address":"[fd00:1122:3344:101::4]:32221","build":"v22.1.9-dirty","started_at":"2024-05-24T16:41:42.706769Z","updated_at":"2024-05-24T19:01:10.944673Z","locality":"","is_available":true,"is_live":true},{"node_id":"5","address":"[fd00:1122:3344:101::3]:32221","sql_address":"[fd00:1122:3344:101::3]:32221","build":"v22.1.9-dirty","started_at":"2024-05-24T16:41:43.079549Z","updated_at":"2024-05-24T19:01:11.326557Z","locality":"","is_available":true,"is_live":true}]}

[smklein]

Nitpick, feel free to tell me to go do this myself, but if we're passing the socketaddr directly to the cockroach-admin service, should we do the same for the cockroachdb service?

(Just feels weird to be asymmetrical with "ip + port" vs "socketaddr" here)

[jgallagher]

Yeah that's fair. I skimmed over this and thought maybe cockroach wanted them as separate args, but it looks like our method script just squishes them back into a socket address. I'll change this to match.

[jgallagher]

Fixed in 7e3ca84 (after doing a bit of rework in 312e11d to clarify where we want an IP vs a socket address)

[smklein]

I'm a little surprised we need this + the build.rs script. I thought we would have needed this to get access to a copy of postgres - this is normally necessary for Diesel, but is it also necessary to invoke the CLI?

[jgallagher]

It's required by the unit tests that want to spin up cockroach. I could maybe change this to a dev-dependency, but then I have to tweak the build script to also only set rpaths when building for tests? Seems fine-ish, although we have that

// NOTE: This file MUST be kept in sync with the other build.rs files in this
// repository.

bit in the copy-paste build.rs that wouldn't be valid anymore.

[smklein]

ahhhh, gotcha, because the unit test scaffolding is making calls with diesel? no objection to this as-is, just was curious

[jgallagher]

I guess so - I didn't really look into it when I saw the rpath failure in CI.

[smklein]

hell yeah, was looking for this test, thrilled it's here