CrYOLO model path caching means changes are not detected during collection

Need to find a way to cache the call (https://github.com/DiamondLightSource/python-murfey/blob/a5e1fa5465023fa8c123e9b9d9d742c4de48b05f/src/murfey/util/processing_params.py#L38) most of the time but clear the cache when needed

Start by reading src/murfey/util/processing_params.py around line 38 and trace how the CrYOLO model path is read during collection. Determine when the cached call must be invalidated so model-path changes are detected while the usual calls remain cached; done means collection observes changed paths without unnecessary repeated work.

索引モデルが issue の本文から書いたものです。