feat: add CPU-aware parallel worker pool for partition merging

Introduce CpuSample to measure process-level CPU efficiency and wall-clock time. Use crossbeam-channel to distribute partition merging tasks to a dynamic worker pool that scales based on CPU utilization, capped at half the available cores. Update diagnostics to track pool usage.

src/obisys/src/lib.rs

@@ -212,6 +212,39 @@ fn rss_to_bytes(ru: &rusage) -> u64 { ru.ru_maxrss as u64 * 1024 }
 // Monotonically increasing counters — negative delta would be a kernel bug.
 fn delta(end: i64, start: i64) -> u64 { (end - start).max(0) as u64 }
 
+// ── CpuSample ─────────────────────────────────────────────────────────────────
+
+/// Snapshot of process-wide CPU time + wall clock at a point in time.
+/// Use [`cpu_efficiency`](Self::cpu_efficiency) to measure the fraction of
+/// available cores used since the snapshot was taken.
+pub struct CpuSample {
+    wall:      Instant,
+    user_secs: f64,
+    sys_secs:  f64,
+}
+
+impl CpuSample {
+    pub fn now() -> Self {
+        let ru = get_rusage();
+        Self {
+            wall:      Instant::now(),
+            user_secs: tv_to_secs(ru.ru_utime),
+            sys_secs:  tv_to_secs(ru.ru_stime),
+        }
+    }
+
+    /// (user_delta + sys_delta) / (wall_delta × n_cores) since this snapshot.
+    /// Returns 0.0 if less than 100 ms have elapsed (too noisy).
+    pub fn cpu_efficiency(&self, n_cores: usize) -> f64 {
+        let ru   = get_rusage();
+        let wall = self.wall.elapsed().as_secs_f64();
+        if wall < 0.1 { return 0.0; }
+        let cpu = (tv_to_secs(ru.ru_utime) - self.user_secs)
+                + (tv_to_secs(ru.ru_stime) - self.sys_secs);
+        cpu / (wall * n_cores as f64)
+    }
+}
+
 // ── public API ────────────────────────────────────────────────────────────────
 
 /// Snapshot taken at the start of a pipeline stage.