DOC: Docs point to incorrect reduce for ma.alltrue and ma.sometrue

Issue with current documentation:

When you run help(np.ma.alltrue) or help(np.ma.sometrue), you get

Help on method reduce in module numpy.ma.core:

reduce(target, axis=0, dtype=None) method of numpy.ma.core._MaskedBinaryOperation instance
    Reduce `target` along the given `axis`.

When you run help on any of the other 50+ ufuncs in the ma module, you get an appropriate doc message. For example help(np.ma.logical_and) yields (the first few lines):

Help on _MaskedBinaryOperation in module numpy.ma.core:

logical_and = <numpy.ma.core._MaskedBinaryOperation object>
    logical_and(x1, x2, /, out=None, *, where=True, casting='same_kind', order='K', dtype=None, subok=True[, signature])
    
    Compute the truth value of x1 AND x2 element-wise.
... it continues....

The issue seems to be the .reduce that appears on just these two functions (see below).
https://github.com/numpy/numpy/blob/98e86d52ed79eb8810960bdcfac11271fc6c5434/numpy/ma/core.py#L1282-L1286

One consequence is that the attributes are different, with ma.alltrue and ma.sometrue being recognized as routines (while none of the others are).

>>>[getattr(np.ma,'logical_and'),getattr(np.ma,'alltrue')]
[<numpy.ma.core._MaskedBinaryOperation at 0x7e0545aa66d0>,
 <bound method _MaskedBinaryOperation.reduce of <numpy.ma.core._MaskedBinaryOperation object at 0x7e0545aa6750>>]
>>>[inspect.isroutine(np.ma.logical_and),inspect.isroutine(np.ma.alltrue)]
[False, True]

I'm hoping to get a script put into the CI build (thanks to @ngoldbaum 's suggestion) that checks to make sure new routines are always appropriately tagged in a .rst file so they get documented on the web.

Idea or request for content:

I have a couple questions:

• How can we handle the ending .reduce at the of the definitions of ma.alltrue and ma.sometrue in a way that properly tags both the same as the rest of the ufuncs? Is that possible (or even wanted).
• Do we want any of these ufuncs showing up in the web documentation as routines of the ma module? They all appear already with help(), but don't appear online.

Start with numpy/ma/core.py at the referenced definitions around lines 1282-1286, then compare the help output for ma.alltrue and ma.sometrue with other masked ufuncs such as ma.logical_and. Review the relevant .rst documentation entries and the proposed CI check in issue #21351; done means the two functions have the intended help and web-documentation classification.