gh-154874: Fix curses.termattrs() returning a negative attribute mask

[mandarwagh9]

termattrs() returns a chtype attribute mask, but GH-134327 routed it through
NoArgReturnIntFunctionBody, which stores the result in an int so it can be compared
against ERR. On a terminal that advertises A_ITALIC — the topmost bit of a 32-bit
mask — that sign-extends, and the mask comes back negative:

>>> curses.termattrs()
-2130771968
>>> curses.term_attrs()          # the same bits, unsigned
2164195328
>>> curses.newwin(1, 1).attrset(curses.termattrs())
OverflowError: can't convert negative value to unsigned int

The value is now returned unsigned, the same way term_attrs(), slk_attr() and
window.getattrs() already do. The cast goes through chtype rather than straight to
unsigned long, matching getattrs(), so no sign extension can happen on a platform
that declares termattrs() as returning an int.

baudrate() is the macro's only other user and really does return a status, so it keeps
the ERR check. The macro gains a short comment saying it is only for such functions —
the same mistake is easy to repeat, and an all-bits-set mask would additionally be
misread as ERR.

Test

TermAttrsTests drives newterm('xterm-256color') over a pseudo-terminal rather than
relying on $TERM, which on CI is usually linux and advertises no italic. Its two
assertions hold on any terminal, so it never skips for lack of an italic-capable one; it
fails without the C change wherever the top bit is actually advertised.

Backport

git branch -r --contains 30dde1ee (the GH-134327 merge) gives main and 3.15 only,
so this needs a 3.15 backport and nothing older. 3.14 is unaffected. The issue carries
3.14 and 3.16 labels as well, but 3.14 predates the regression.

Notes

• gh-154874: Fix the sign of curses.termattrs() #154875 proposes the same one-line fix; I opened this independently and only saw it
  afterwards. Happy for whichever is more useful to maintainers to land — closing this
  one is completely fine.
• I could not build or run _curses locally (Windows), so the test here is unverified
  by me and I am relying on CI to run it. Worth a careful look on that account.
• Prepared with AI assistance (Claude); the reasoning and the diff have been reviewed by
  me, but please weigh the point above accordingly.

• Issue: curses.termattrs() returns a negative value on italic terminals #154874

[Copilot]

Copilot was unable to review this pull request because the user who requested the review has reached their quota limit.