Describe the issue:

The output of str(np.dtypes.StringDType()) is 'StringDType()'.
The output of np.dtypes.StringDType().str is '|T16'.

I think that the .str output is quite uninformative, and it would be much better as 'StringDType()' too.
Namely, I don't think that the final user should be exposed to the word-size, which is not representative of the consumed memory, is an implementation detail, and (to my understanding) changes between 32 and 64 bit architectures.

Reproduce the code example:

>>> np.dtypes.StringDType().str
|T16

Python and NumPy Versions:

NumPy 2.2.4

Context for the issue:

h5py/h5py#2557 (comment)

versioned_hdf5, like many other wrappers around NumPy dtypes, needs to pretty-print metadata information in their __repr__ methods. We had to special-case around this.