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.

Change final_dims_fake back after reoder array because if retrieve = F, it will be used in attributes later

Move adjusted split dim order part to after 'Detect dimension' print. Then reorder dimension back to the user-required order after getting array.