`convert-s2-optimized` fails due to chunk misalignment unless `use-sharding` is `True`

[d-v-b]

unless we enter this branch:

data-model/src/eopf_geozarr/s2_optimization/s2_multiscale.py

Lines 410 to 411 in 41c3490

	if self.enable_sharding:
	dataset = self._rechunk_dataset_for_encoding(dataset, encoding)

, there will be a chunk misalignment and xarray will refuse to write Zarr data. So we should probably rechunk whether sharding is used or not.

This will be fixed in #75