test: retire stale connections scenario tests (command removed in #221)

[eddietejeda]

What

Retire two scenario tests that call the hotdata connections command removed in #221 ("scope all commands to active database; remove connections command").

Background

#221 deleted the connections command but left these tests behind (and never updated the server-side scenario manifest). They've been silently broken since:

• tests/connections_read.rs — calls the removed connections list / connections <id>. It exists only to satisfy scenario-parity's file-existence check and skips in CI (HOTDATA_SDK_TEST_CONNECTION_ID is unset), so nobody noticed it was dead. The scenario has no current CLI surface (connections are resolved internally now). Deleted.
• tests/auth_unknown_workspace.rs — a security scenario: a fabricated workspace id must be rejected with no cross-workspace leak. It also called connections list, so it exited non-zero at arg-parse — i.e. it was passing for the wrong reason and never actually reached the server to verify the rejection. Switched to databases list, a real workspace-scoped listing that hits the gateway with the fabricated workspace id, restoring genuine coverage. (The scenario's ops: ["*"] allows any op.)

Dependency / merge order

Companion PR hotdata-dev/www.hotdata.dev#290 adds cli to optional_for on the connections_read scenario. That must merge first — until it does, scenario-parity fetches the live manifest (which still requires the CLI to have tests/connections_read.rs) and this PR's scenario-parity check will be red. Once #290 is on main, re-run and it goes green.

Verification

• cargo test + cargo fmt --check pass locally; auth_unknown_workspace compiles and skips cleanly without creds.
• No remaining references to connections list outside an explanatory comment.

[claude]

nit: (not blocking) this was the only caller of skip_if_no_connection!, so deleting it leaves the whole connection-id scaffolding dead:

• tests/common/mod.rs:265-292 — the skip_if_no_connection! macro (won't warn, since #[macro_export] macros are exempt from unused_macros)
• tests/common/mod.rs:48,70 — TestEnv::connection_id and its load_env wiring (silenced by the crate-level #![allow(dead_code)])
• .github/workflows/integration-tests.yml:98 — HOTDATA_SDK_TEST_CONNECTION_ID passed to the integration job

Since connections are resolved internally now and no scenario is expected to re-add a connections CLI surface, dropping these in the same PR keeps the harness honest — otherwise the next reader sees a supported connection-gated path that nothing uses. Fine to leave if you'd rather keep it as a hook for a future scenario.

[claude]

Verified the substantive claim: -w is global = true on Commands::Databases (src/cli.rs:52-58) and resolve_workspace (src/main.rs:74-78) returns an explicit --workspace-id verbatim with no client-side validation, so the fabricated id does reach the gateway as X-Workspace-Id. databases::list exits non-zero via e.exit() on API error (src/commands/databases.rs:943), so the non-zero assertion is real coverage now rather than an arg-parse artifact.

One non-blocking nit inline.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.

📢 Thoughts on this report? Let us know!