This PR makes a ChunkExecutor which is agnostic to the recording and applies to any chunkable BaseExtractor (which defines get_sample_size(), get_num_samples(), get_num_segments()

This is because we are working on a new "spikeinterface" for calcium imaging, and this refactor makes it suuuuper straightforward to use the same parallelization machinery for Imaging objects (which inherit from the BaseExtractor)

Update 29/11/25

Following up on trying to make SI tools as general as possible for its calcium imaging "clone" (photon-mosaic) I made some more dramatic refactoring:

• implemented a ChunkableMixin abstract mixin that implements all method needed by the ChunkExecutor
  • several BaseRecording specific methods (e.g., write_binary/write_memory and more) are now in the chunkable_tools and applicable to any ChunkableMixin object
  • the ChunkableMixin has a general get_data function, so it can be used by many derived classes (e.g., get_traces for recordings, get_videos/series for imaging)
  • another useful function to centralize things is the get_shape(segment_index), which returns the buffer shape by segment
• refactored "segments": I that we didn't have any proper structure for segments: recording objects had _recording_segments and sorting objects _sorting_segments. Each class then implemented its own add_* function. Since segments are a BaseExtractor concept, I centralized the definition (self.segments: List[BaseSegment] = []), added a self.segments property, and a centralized add_segment function.

@samuelgarcia @chrishalcrow @yger @h-mayorquin @JoeZiminski @zm711

Let me know what you think!!!