ComputeWaveforms and ComputeTemplates: int() truncation causes inconsistent sample counts across slightly different sampling rates

[Antorhythms]

In analyzer_extension_core.py, the nbefore and nafter properties in ComputeWaveforms and ComputeTemplates use int() (truncation) to convert millisecond parameters to sample counts:

@property
def nbefore(self):    
	return int(self.params["ms_before"] * self.sorting_analyzer.sampling_frequency / 1000.0)

This causes a ±1 sample inconsistency when the sampling rate deviates slightly from a round number. For example, with ms_before=2.0:
fs = 29999 Hz → int(59.998) = 59
fs = 30000 Hz → int(60.0) = 60

This means waveform arrays have different shapes and different center indices across datasets with nearly identical sampling rates.

Proposed fix:
Replace int() with round() in the nbefore/nafter properties of both ComputeWaveforms (lines 163, 167) and ComputeTemplates (lines 526, 532).

[alejoe91]

Hi @Antorhythms

Would you like to make a PR for this issue? :)