Measure dropped samples from stream timestamps instead of hiding them

Overflows were counted but the samples the driver dropped were simply
missing, so a 5 s capture with 300 overflows still read as a clean 5.000 s
file. Each read's hardware timestamp is now compared with where the
previous read ended; gaps are sized exactly, zero-filled by default so the
file timeline stays true, and reported as completed_with_gaps with a
warning, samples_dropped and a gap list. This also catches drops the
plugin never flags as overflows (libusb transfer timeouts).

Timestamps are differenced before rounding: rounding each absolute
timestamp to a sample index flips by +/-1 near half-sample boundaries
and produced phantom 1-sample gaps.

- stream buffers 128 x 64K samples (~280 ms at 30 Msps, was ~4 ms)
- reject out-of-range gain, frequency, rate and bandwidth instead of
  letting the plugin clamp silently; default bandwidth clamps to range
- spectrum returns a max-hold decimated summary (256 points) plus
  noise_floor_db; full arrays were 139 KB at fft_size 4096
- capture_iq background=True with stream_status polling, so
  stop_capture is reachable; device-changing tools refuse while it runs
- OSError returned in the error shape; status drops "unknown" fields
- README: gap accounting, keep captures off tmpfs; captures/ ignored

tests/test_capture.py

@@ -34,17 +34,6 @@ def test_cs16_file_is_raw_q11(tmp_path):
     np.testing.assert_allclose(x, [1.0 - 0.5j, 0.25j])
 
 
-def test_components_and_clip_level_agree_across_formats():
-    q11 = np.array([2047, -100, 50, -2048], dtype=np.int16)
-    cf32 = (q11[0::2] + 1j * q11[1::2]).astype(np.complex64) / device.CS16_FULL_SCALE
-
-    a = capture._components(q11, "CS16")
-    b = capture._components(cf32, "CF32")
-
-    np.testing.assert_allclose(a, b)
-    assert np.count_nonzero(a.max(axis=1) >= capture._CLIP_LEVEL) == 2
-
-
 class _Range:
     def __init__(self, lo, hi):
         self.lo, self.hi = lo, hi
@@ -63,6 +52,36 @@ def test_range_span_covers_every_range():
     assert device.range_span([]) is None
 
 
+def test_chunk_stats_match_components():
+    q11 = np.array([2047, -100, 50, -2048, 10, 10], dtype=np.int16)
+    cf32 = (q11[0::2] + 1j * q11[1::2]).astype(np.complex64) / device.CS16_FULL_SCALE
+
+    for chunk, fmt in ((q11, "CS16"), (cf32, "CF32")):
+        peak, clipped = capture._chunk_stats(chunk, fmt)
+        assert peak == pytest.approx(1.0)
+        assert clipped == 2
+
+
+def test_decimate_max_hold_keeps_narrow_peak():
+    f = list(range(4096))
+    p = [-80.0] * 4096
+    p[1234] = -20.0
+
+    df, dp = capture.decimate_max_hold(f, p, 256)
+
+    assert len(dp) == 256
+    assert max(dp) == -20.0
+    assert capture.decimate_max_hold(f[:100], p[:100], 256) == (f[:100], p[:100])
+
+
+def test_check_range_rejects_instead_of_clamping():
+    device.check_range("rxvga1 gain", 30, _Range(5, 30), "dB")
+    with pytest.raises(RuntimeError, match="outside the supported range 5 to 30 dB"):
+        device.check_range("rxvga1 gain", 99, _Range(5, 30), "dB")
+    with pytest.raises(RuntimeError, match="80000 to 4e\\+07"):
+        device.check_range("sample rate", 61.44e6, [_Range(80e3, 10e6), _Range(20e6, 40e6)], "Hz")
+
+
 def test_alloc_buffer_matches_stream_format():
     assert device.alloc_buffer(10, "CS16").dtype == np.int16
     assert device.alloc_buffer(10, "CS16").size == 20