fix: preserve time_resolution in crop_frequency() and crop_time() (#54)

Both crop methods were not passing time_resolution to new SpectrogramData
instances, causing them to fall back to DEFAULT_TIME_RESOLUTION (0.01508s)
instead of the accurate value calculated from database duration (e.g. 0.015103s).

This small discrepancy accumulates over long recordings, causing significant
timing misalignment. For a 2192-second recording, the error reached ~0.15s.

Now both methods preserve the parent spectrogram's time_resolution, ensuring
accurate timing throughout the crop chain.

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-authored-by: Claude <noreply@anthropic.com>

Author: jon-myers

idtap/spectrogram.py

@@ -243,7 +243,8 @@ def crop_frequency(self, min_hz: Optional[float] = None,
             cropped_data,
             self.audio_id,
             freq_range=(freq_bins[min_idx], freq_bins[max_idx - 1] if max_idx > min_idx else freq_bins[min_idx]),
-            bins_per_octave=self.bins_per_octave
+            bins_per_octave=self.bins_per_octave,
+            time_resolution=self._time_resolution
         )
 
     def crop_time(self, start_time: Optional[float] = None,
@@ -277,7 +278,8 @@ def crop_time(self, start_time: Optional[float] = None,
             cropped_data,
             self.audio_id,
             freq_range=self.freq_range,
-            bins_per_octave=self.bins_per_octave
+            bins_per_octave=self.bins_per_octave,
+            time_resolution=self._time_resolution
         )
 
     def get_extent(self) -> List[float]: