If the name of return_vars is the synonmin of inner dim name, change it back to inner dim name. Unit test added.

Add new params 'largest_dims_length' to determine if Start() should examine all the files to get the dim or not

Add one 'if' condition to ensure the first file of each dataset exists. If not, metadata is assigned to NULL.

Bugfix for Start() parameter 'path_glob_permissive' and write the explanation in FAQ and Start() header.

Remove the indices_chunk part to the first, so if the array doesn't need reorganization, the new development is skipped.

Add additional data array rearrangement if merge_across_dim + split_multiselected_dims. It corrects the displaced values when the values in the same file are called more than 1 time.