curses window.in_wstr() returns garbage when reading no characters

[fedonman]

Bug description:

window.in_wstr(y, x, 0) reads uninitialized heap memory. winnwstr() only writes the
terminating null when it stored at least one character, but the buffer comes from
PyMem_New and the result is read with PyUnicode_FromWideChar(buf, -1), which scans for
that null (Modules/_cursesmodule.c:3853, :3869). The value differs on every run:

in_wstr(0, 0, 0)       -> ValueError: character U+e7180240 is not in range [U+0000; U+10ffff]
in_wstr(0, 0, 0)       -> ValueError: character U+20b80240 is not in range [U+0000; U+10ffff]

instr(0, 0, 0) and in_wchstr(0, 0, 0) correctly return b'' and
curses.complexstr([]).

gh-152503 fixed this for in_wch(), in_wchstr() and getbkgrnd(); in_wstr() was
missed, and unlike those it reproduces on stock ncursesw. It is new in 3.16 (gh-151757), so
no released version is affected.

CPython versions tested on:

CPython main branch

Operating systems tested on:

Linux

Linked PRs

• gh-154781: Zero the in_wstr() buffer in curses #154782