trunction cause duration < 0

[johnchienbronci]

@pzelasko Hi
As long as I add noise, and only read the segmentation audio file with (offset + duration)
result: new_duration< 0 error

File "/usr/local/lib/python3.10/dist-packages/lhotse/dataset/sampling/dynamic_bucketing.py", line 299, in _next_batch                                                                       
    batch = next(self.cuts_iter)                                      
  File "/usr/local/lib/python3.10/dist-packages/lhotse/dataset/sampling/dynamic_bucketing.py", line 588, in __iter__                                                                          
    self._collect_cuts_in_buckets(self.buffer_size)                                            
  File "/usr/local/lib/python3.10/dist-packages/lhotse/dataset/sampling/dynamic_bucketing.py", line 768, in _collect_cuts_in_buckets                                                          
    cuts = next(self.cuts_iter)                                                               
  File "/usr/local/lib/python3.10/dist-packages/lhotse/dataset/sampling/dynamic.py", line 377, in __iter__                                                                                    
    for item in self.iterator:                                                                                             
  File "/usr/local/lib/python3.10/dist-packages/lhotse/cut/set.py", line 2554, in __iter__                                                                                                    
    yield from self.cuts                                                                                            
  File "/usr/local/lib/python3.10/dist-packages/lhotse/cut/set.py", line 2554, in __iter__                                                                                                    
    yield from self.cuts                                                                                           
  File "/usr/local/lib/python3.10/dist-packages/lhotse/cut/set.py", line 3587, in __iter__                                                                                                    
    to_mix = self._maybe_truncate_cut(to_mix, target_mixed_duration, rng)                                                                                                                     
  File "/usr/local/lib/python3.10/dist-packages/lhotse/cut/set.py", line 3629, in _maybe_truncate_cut                                                                                         
    cut = cut.truncate(                    
  File "/usr/local/lib/python3.10/dist-packages/lhotse/cut/data.py", line 520, in truncate                                                                                                    
    assert new_duration > 0.0, f"new_duration={new_duration}, until={until}, offset={offset}, self.start={self.start}, new_start={new_start}, self.duration={self.duration}"         
AssertionError: new_duration=-2.91, until=0.5120625, offset=3.4220819912490628, self.start=0.0, new_start=3.4220625, self.duration=5.705125        

read audio: MonoCut(id='xxx', start=0.0, duration=12.46, channel=0, supervisions=[], features=None, recording=Recording(id='xxx', sources=[AudioSource(type='file', channels=[0], source='xxx')], sampling_rate=16000, num_samples=199360, duration=12.46, channel_ids=[0], transforms=None), custom=None): duration=12.46, offset=15.32

truncate:
add_durations: tot_num_samples:  245120 sampling_rate:  16000
add_durations: tot_num_samples:  444480 sampling_rate: 16000  
add_durations: tot_num_samples:  199360 sampling_rate:  16000                                                  
self.start=0.0, offset=15.32, duration=12.46                      
new_start=15.32, until=27.78, new_duration=12.46                
add_durations: tot_num_samples:  245120 sampling_rate:  16000                                                    
add_durations: tot_num_samples:  -45760 sampling_rate:  16000                                              
duration_past_end:  15.32 new_duration: -2.86              
_create_cut result: : MonoCut(id='xxx-001532-001246', start=15.32, duration=-2.86, channel=0,supervisions=[], features=None, recording=Recording(id='xxx', sources=[AudioSource(type='file', channels=[0], source='xxx')], sampling_rate=16000, num_samples=199360, duration=12.46, channel_ids=[0], transforms=None), custom=None)  

truncate:
add_durations: tot_num_samples:  54753 sampling_rate:  16000                                                 
add_durations: tot_num_samples:  8193 sampling_rate:  16000  
add_durations: tot_num_samples:  -46560 sampling_rate:  16000
self.start=0.0, offset=3.4220819912490628, duration=-2.91    
new_start=3.4220625, until=0.5120625, new_duration=-2.91