DOC: Array interface protocol does not recognize bit field typestr and could mention AttributeError use

Describe the issue:

I am trying to implement the array interface protocol for Polars Series. Polars represents its boolean arrays (and validity buffers) as bit-packed booleans, e.g. 8 booleans per byte. I figured I would need to specify a bit field (e.g. t type) to do this, however, it seems numpy does not recognize any way of defining a bit field. I've tried various combinations (with descr too) to get this to work, but I can't figure it out.

Could it be that NumPy does not handle bit field types correctly?

Reproduce the code example:

import numpy as np


class wrapper:
    pass


interface = {
    "shape": (3,),
    "typestr": "|t1",
    "version": 3,
}
wrapper.__array_interface__ = interface

result = np.array(wrapper, copy=False)

Error message:

Traceback (most recent call last):
  File "/home/stijn/code/polars/py-polars/repro.py", line 19, in <module>
    result = np.array(wrapper, copy=False)
             ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
TypeError: data type '|t1' not understood

Python and NumPy Versions:

1.26.3
3.12.0 (main, Nov 16 2023, 08:47:32) [GCC 11.4.0]

Runtime Environment:

[{'numpy_version': '1.26.3',
'python': '3.12.0 (main, Nov 16 2023, 08:47:32) [GCC 11.4.0]',
'uname': uname_result(system='Linux', node='frosty', release='6.1.0-1029-oem', version='#29-Ubuntu SMP PREEMPT_DYNAMIC Tue Jan 9 21:07:34 UTC 2024', machine='x86_64')},
{'simd_extensions': {'baseline': ['SSE', 'SSE2', 'SSE3'],
'found': ['SSSE3',
'SSE41',
'POPCNT',
'SSE42',
'AVX',
'F16C',
'FMA3',
'AVX2'],
'not_found': ['AVX512F',
'AVX512CD',
'AVX512_KNL',
'AVX512_KNM',
'AVX512_SKX',
'AVX512_CLX',
'AVX512_CNL',
'AVX512_ICL']}},
{'architecture': 'Prescott',
'filepath': '/home/stijn/code/polars/.venv/lib/python3.12/site-packages/numpy.libs/libopenblas64_p-r0-0cf96a72.3.23.dev.so',
'internal_api': 'openblas',
'num_threads': 12,
'prefix': 'libopenblas',
'threading_layer': 'pthreads',
'user_api': 'blas',
'version': '0.3.23.dev'}]
None

Context for the issue:

I am trying to improve interoperability between the Polars dataframe library and NumPy.

Start by running the provided array_interface reproduction and tracing NumPy's parsing of the "typestr" field. Determine whether bit-packed boolean types are supported and how the reported AttributeError usage should be addressed; done requires an agreed behavior with regression coverage or documentation.