Dataset MUST NOT contain duplicate obs, by raw counts

[jahilton]

Currently, curators perform a QA check to see if any two obs in a Dataset have the same raw count value for every feature
code in evaluate_dup_counts() but can be improved - at least one gap exists where the current check assumes that the csr matrix underlying arrays represent the raw matrix in an ordered fashion, but this is not the case. The indices array can be unsorted and potentially in a different order per row; duplicates are also possible in this scenario.

Suggest the table in X (Matrix Layers) be updated in the "raw" required? column.
For the two scRNA-seq rows add Each cell MUST NOT have the same value for every feature as any other cell.
For the Visium row add If the obs['in_tissue'] value is 1, then each cell MUST NOT have the same value for every feature as any other cell

We have one existing case in the corpus. The contributors have been given a deadline of Dec 12 to resolve the issue (#528).