When rechunking by the current chunk size, name should not change.
Add a __dask_tokenize__ method for ReprObject so that this behaviour is present
when DataArrays are converted to temporary Datasets and back.

…chunk-unique-token

* 'chunk-unique-token' of github.com:dcherian/xarray:
  remove more computes.

* upstream/master:
  Fix map_blocks HLG layering (pydata#3598)
  Silence sphinx warnings: Round 2 (pydata#3592)
  2x~5x speed up for isel() in most cases (pydata#3533)
  remove xarray again (pydata#3591)
  fix plotting with transposed nondim coords. (pydata#3441)
  make coarsen reductions consistent with reductions on other classes (pydata#3500)
  Resolve the version issues on RTD (pydata#3589)
  Add bottleneck & rasterio git tip to upstream-dev CI (pydata#3585)

…oken

* 'master' of github.com:pydata/xarray:
  Add nanmedian for dask arrays (pydata#3604)
  added pyinterp to related projects (pydata#3655)
  Allow incomplete hypercubes in combine_by_coords (pydata#3649)
  concat keeps attrs from first variable. (pydata#3637)
  Extend DatetimeAccessor properties and support `.dt` accessor for Timedelta (pydata#3612)
  update readthedocs.yml (pydata#3639)
  silence sphinx warnings round 3 (pydata#3602)
  Fix/quantile wrong errmsg (pydata#3635)
  Provide shape info in shape mismatch error. (pydata#3619)
  Minor doc fixes (pydata#3615)
  Respect user-specified coordinates attribute. (pydata#3487)
  Add Facetgrid.row_labels & Facetgrid.col_labels (pydata#3597)
  Fix pint integration tests (pydata#3600)
  Minor fix to combine_by_coords to allow for the combination of CFTimeIndexes separated by large time intervals (pydata#3543)

Co-Authored-By: crusaderky <crusaderky@gmail.com>