cosmicray_lacosmic returns float32.

Cannot control the precision of the array returned by cosmicray_lacosmic. Always get float32.

The issue identifies cosmicray_lacosmic as the entry point; begin by locating that function and reproducing its current float32 return with arrays of different precision. Determine how the result dtype is selected and add coverage for the requested control; done means callers can choose the returned array precision rather than always receiving float32.

Written by the indexing model from the issue text.