[BUG] Raise Erro when the max number of counts table more than int32

[byemaxx]

If my count table include number more than int32 (2147483647), it will raise Erro as follows:

in DeseqDataSet.__init__(self, adata, counts, clinical, design_factors, ref_level, min_mu, min_disp, max_disp, refit_cooks, min_replicates, beta_tol, n_cpus, batch_size, joblib_verbosity)
    193     self.__dict__.update(adata.__dict__)
    194 elif counts is not None and clinical is not None:
    195     # Test counts before going further
--> 196     test_valid_counts(counts)
    197     super().__init__(X=counts.astype(int), obs=clinical)
    198 else:
...
    135     raise ValueError("The count matrix should only contain integers.")
    136 if (counts < 0).any().any():
--> 137     raise ValueError("The count matrix should only contain non-negative values.")

ValueError: The count matrix should only contain non-negative values.

Expected behavior
A temporary solution is to zoom in data before creating dds, as follows:

if counts_df.max().max() > 2147483647:
    times = counts_df.max().max() / 2147483647
    divide = int(times) + 1
    counts_df = counts_df / divide