gh-70546: Document mmap object's __iter__ behavior that differs from that of bytearray.

[ztane]

#70546 reports that mmap object __iter__ behavior dffers from that of bytearray. This pr adds a notice that states that mmap objects have __iter__ that differs from that of bytearrays and that you can wrap the object inside a memoryview to work around the difference.

• Issue: mmap.mmap.__iter__ is broken (yields bytes instead of ints) #70546

[ztane]

If this PR is merged, I think the referenced issue could be also closed (I cannot do it because it still belongs to a mannequin ;)

[sergey-miryanov]

I believe this shouldn't skip news, because it is a user-facing update.
Also, maybe it is worth to add an example of usage memoryview to the note?

[sergey-miryanov]

@serhiy-storchaka Could you please take a look?

[serhiy-storchaka]

It is not only iterating (mmap does not even implement explicit __iter__), but subscribing by an index which returns a bytes object of length 1. And assignment requires a bytes object of length 1, not integer.

To add insult to injury, the code is self-contradictory -- so PyObject_GetItem() returns an integer, but PySequence_GetItem() returns a bytes object of length 1. This is perhaps not visible from Python side.

The documentation should mention all four operations which are different between mmap and other bytes-like objects (bytes, bytearray, array.array, memview): subscribing by an index, assigning by an index, iteration and the containment check.