gh-133489: Document random.randbytes bounds.

[StanFromIreland]

I believe the upper bound is not platform dependent.

Closes #133489

• Issue: Possible unnecessary OverflowError in random.getrandbits #133489

---

📚 Documentation preview 📚: https://cpython-previews--133529.org.readthedocs.build/

[skirpichev]

I believe the upper bound is not platform dependent.

Mentioned limit holds just because n*8 in randbytes(n) should fit int's size. Typically it's 4 bytes, thus your 2 ** 28. But I'm not sure it's platform-independent.

[rhettinger]

This isn't something we normally document. Consider len(range(2**65)) for example. To some degree, OverflowErrors speak for themselves. Also, putting a note like this in the docs doesn't really help the reader.

Also, this isn't specific to randbytes. This happens upstream with the random.getrandbits(8 * 2 ** 28) call.