Windows: encode and decode text as UTF-8 instead of trusting the locale

setlocale(LC_ALL, "") maps to the ANSI code page on Windows -- 936 (GBK)
on a Chinese system -- so wcrtomb() produces GBK where the protocol wants
UTF-8, and mbrtowc() reads UTF-8 as GBK. Typed Chinese therefore reached
the model as mojibake, and once the buffer held correct UTF-8 the input
line rendered as mojibake instead. Windows' C runtime has no ".UTF-8"
locale to switch to, so the conversion has to be explicit.

os_compat.[ch]
  * add os_utf8_encode() and os_utf8_decode(), locale-independent, with
    range and overlong-form checks
src/tui.c
  * input_put_wc(), the command-palette filter and the ask_user buffer use
    os_utf8_encode() instead of wcrtomb()
src/textsel.c
  * cst_utf8_decode() uses os_utf8_decode() instead of mbrtowc()

include/os_compat.h

@@ -111,6 +111,29 @@ void os_localtime(const time_t *t, struct tm *out);
 int os_getpid(void);
 int os_chmod_private(const char *path); /* best-effort 0600; no-op/ok on Win */
 
+/* --- UTF-8 --------------------------------------------------------------
+ * Encode one code point as UTF-8, independent of the C locale.
+ *
+ * wcrtomb() follows setlocale(). On Linux the environment normally selects a
+ * UTF-8 locale, but on Windows the C runtime maps setlocale(LC_ALL, "") to the
+ * ANSI code page -- 936 (GBK) on a Chinese system -- so wcrtomb() hands back
+ * GBK bytes where the protocol requires UTF-8, and the model receives
+ * mojibake. Windows' C runtime has no ".UTF-8" locale to switch to.
+ *
+ * Returns the number of bytes written (1..4), or 0 if the code point cannot be
+ * encoded. Surrogate halves (as they arrive on 16-bit wchar_t platforms for
+ * non-BMP code points) return 0 and are skipped. */
+int os_utf8_encode(unsigned long cp, char out[4]);
+
+/* Decode one UTF-8 sequence from s (at most avail bytes).
+ * Returns the number of bytes consumed (1..4) and stores the code point in
+ * *cp, or 0 if the bytes are not a valid, complete, non-overlong sequence.
+ *
+ * The counterpart of the encoder above, and needed for the same reason:
+ * mbrtowc() also follows setlocale(), so on Windows it would decode UTF-8
+ * text as GBK and the screen would show mojibake. */
+int os_utf8_decode(const char *s, size_t avail, unsigned long *cp);
+
 /* --- libc gaps on Windows ------------------------------------------------
  * MinGW-w64 provides neither strcasestr() (a GNU extension) nor
  * setenv()/unsetenv() (POSIX): they are simply absent from its headers.