hive-c0re: spawn the auto-update and knowledge git calls through git_command

Three git spawns in workers/auto_update.rs and workers/knowledge.rs used
a bare tokio Command, so they ignored HYPERHIVE_GIT and, when the job
queue's deadline dropped their future (the knowledge pull has a 5 minute
one), the git child was left running instead of killed.

They now use lifecycle::git_command(), as the other git calls in
knowledge.rs already do.

hive-c0re/src/workers/auto_update.rs

@@ -51,7 +51,7 @@ pub fn current_flake_rev(hyperhive_flake: &str) -> Option<String> {
 /// starved the runtime and stalled the per-agent sockets.
 pub async fn agent_config_pending(name: &str, deployed_sha: Option<&str>) -> bool {
     let applied = crate::paths::applied_dir(name);
-    let applied_head = tokio::process::Command::new("git")
+    let applied_head = crate::lifecycle::git_command()
         .args(["-C", &applied.to_string_lossy(), "rev-parse", "HEAD"])
         .output()
         .await

hive-c0re/src/workers/knowledge.rs

@@ -72,7 +72,7 @@ async fn ensure_local_clone(auth: Option<&str>) -> Result<()> {
 /// failure (used only to detect whether a pull actually moved `HEAD` —
 /// worth skipping the broadcast over, not worth failing the pull for).
 async fn head_sha() -> Option<String> {
-    let out = tokio::process::Command::new("git")
+    let out = crate::lifecycle::git_command()
         .args(["-C", LOCAL_DIR, "rev-parse", "HEAD"])
         .output()
         .await
@@ -186,7 +186,7 @@ pub async fn pull(coord: &Coordinator) -> Result<()> {
 /// summary to every live agent. Best-effort — a diff or send failure is
 /// logged, never propagated (the pull itself already succeeded).
 async fn broadcast_change(coord: &Coordinator, before: &str, after: &str) {
-    let diff = tokio::process::Command::new("git")
+    let diff = crate::lifecycle::git_command()
         .args([
             "-C",
             LOCAL_DIR,