vLLM priority scheduling for agents

Thread request priority through the API call chain to vLLM's
priority scheduler. Lower value = higher priority, with preemption.

Priority is set per-agent in the .agent header:
- interactive (runner): 0 (default, highest)
- surface-observe: 1 (near-realtime, watches conversation)
- all other agents: 10 (batch, default if not specified)

Requires vLLM started with --scheduling-policy priority.

Co-Authored-By: Proof of Concept <poc@bcachefs.org>

src/subconscious/api.rs

@@ -35,6 +35,7 @@ pub async fn call_api_with_tools(
     prompts: &[String],
     phases: &[String],
     temperature: Option<f32>,
+    priority: i32,
     tools: &[String],
     bail_fn: Option<&(dyn Fn(usize) -> Result<(), String> + Sync)>,
     log: &dyn Fn(&str),
@@ -82,6 +83,7 @@ pub async fn call_api_with_tools(
                 &ui_tx,
                 &reasoning,
                 temperature,
+                Some(priority),
             ).await {
                 Ok((msg, usage)) => {
                     msg_opt = Some(msg);
@@ -233,6 +235,7 @@ pub fn call_api_with_tools_sync(
     prompts: &[String],
     phases: &[String],
     temperature: Option<f32>,
+    priority: i32,
     tools: &[String],
     bail_fn: Option<&(dyn Fn(usize) -> Result<(), String> + Sync)>,
     log: &(dyn Fn(&str) + Sync),
@@ -244,7 +247,7 @@ pub fn call_api_with_tools_sync(
                 .build()
                 .map_err(|e| format!("tokio runtime: {}", e))?;
             rt.block_on(
-                    call_api_with_tools(agent, prompts, phases, temperature, tools, bail_fn, log)
+                    call_api_with_tools(agent, prompts, phases, temperature, priority, tools, bail_fn, log)
             )
         }).join().unwrap()
     })