refactor: remove threading from metrics

The server is pure asyncio (single-threaded). The threading.Lock
was never contended. Use a float accumulator in _human_bytes to
avoid the int-to-float type: ignore.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

src/s5p/metrics.py

@@ -2,19 +2,17 @@
 
 from __future__ import annotations
 
-import threading
 import time
 
 
 class Metrics:
-    """Thread-safe connection metrics.
+    """Connection metrics.
 
-    Counters use a lock for consistency but tolerate minor races
-    on hot paths (bytes_in/bytes_out) for performance.
+    All access runs on the single asyncio event-loop thread,
+    so no locking is needed.
     """
 
     def __init__(self) -> None:
-        self._lock = threading.Lock()
         self.connections: int = 0
         self.success: int = 0
         self.failed: int = 0
@@ -26,7 +24,6 @@ class Metrics:
 
     def summary(self) -> str:
         """One-line log-friendly summary."""
-        with self._lock:
         uptime = time.monotonic() - self.started
         h, rem = divmod(int(uptime), 3600)
         m, s = divmod(rem, 60)
@@ -40,10 +37,11 @@ class Metrics:
 
 def _human_bytes(n: int) -> str:
     """Format byte count in human-readable form."""
+    value = float(n)
     for unit in ("B", "K", "M", "G", "T"):
-        if abs(n) < 1024:
+        if abs(value) < 1024:
             if unit == "B":
-                return f"{n}{unit}"
-            return f"{n:.1f}{unit}"
-        n /= 1024  # type: ignore[assignment]
-    return f"{n:.1f}P"
+                return f"{int(value)}{unit}"
+            return f"{value:.1f}{unit}"
+        value /= 1024
+    return f"{value:.1f}P"