Dask Graph Cleanup needed after Rechunking in _reproject_dispatcher()

Dask rechunking within [`_reproject_dispatcher()`](https://github.com/astropy/reproject/blob/c4c7d0ab0a181e0b47ee0a53e30d84b725ab3af2/reproject/common.py#L55) is likely responsible for the function's memory problem down the line with `da.map_blocks()` reprojecting the blocks. Without minimizing/cleaning the dask graph pre-emptively, e.g., have it computed first, the graph can become unmanageably large quickly during the reprojection.

https://github.com/astropy/reproject/blob/c4c7d0ab0a181e0b47ee0a53e30d84b725ab3af2/reproject/common.py#L295

Inspect reproject/common.py, especially _reproject_dispatcher() near the linked lines and its da.map_blocks() call. Reproduce the memory or graph-growth problem around rechunking, then determine a cleanup or computation point that keeps the graph manageable while preserving reprojection behavior.

Written by the indexing model from the issue text.