BPM tap button averages the whole session, not the last 8 taps

TapTempo kept a rolling window of 8 taps, so a long tapping session
chased the most recent taps instead of converging on the overall tempo
(trav noticed the estimate never settled). Every tap since the session
started now counts; a >2.5s gap still begins a fresh session.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

tests/test_tap_tempo.py

@@ -26,18 +26,27 @@ def test_long_gap_resets_series():
     assert tempo.tap(6.1) == 100  # 0.6s interval
 
 
-def test_window_uses_recent_taps_only():
+def test_long_session_converges_exactly():
     tempo = TapTempo()
-    # 4 slow taps (1s = 60bpm) then 8 fast taps (0.25s = 240bpm);
-    # only the last 8 taps count, so the slow ones age out entirely
-    t = 0.0
-    for _ in range(4):
-        tempo.tap(t)
-        t += 1.0
-    for _ in range(8):
-        tempo.tap(t)
-        t += 0.25
-    assert tempo.bpm() == 240
+    bpm = None
+    for i in range(30):
+        bpm = tempo.tap(i * 0.5)  # 30 steady taps at 120 bpm
+    assert bpm == 120
+
+
+def test_whole_session_is_averaged():
+    tempo = TapTempo()
+    # One sloppy 1.0s interval up front, then 20 taps at 0.5s. All 21 taps
+    # count (no rolling window), so the estimate is the overall average
+    # 10.5s / 20 intervals = 0.525s -> 114 bpm, not the recent-taps 120.
+    tempo.tap(0.0)
+    t = 1.0
+    bpm = None
+    for _ in range(20):
+        bpm = tempo.tap(t)
+        t += 0.5
+    assert tempo.tap_count == 21
+    assert bpm == 114
 
 
 def test_reset():