FIX/CHANGE: wide internal encoding was not used for strings containing chars in range >= 80h and <= FFh

There is no suitable test for it, as there is no function returning internal char size, but with `DUMP` function enabled, it was clearly visible, that before this change:

```
>> dump to-string #{A3}
=> Series #033A8EF0 "-": wide: 1  size: 8      bias: 0 tail: 1 rest: 8 flags: #1
03031670: A300                                ..

```
Notice that there is: `wide: 1` in the output above.

Now with the same code it is 2:
```
>> dump to-string #{A3}
=> Series #7C18FAF0 "-": wide: 2  size: 8      bias: 0 tail: 1 rest: 4 flags: #2
7E0C4CD8: A3000000                            ....
```

Memory consumption in such a cases may be higher, but it simplifies other manipulation as series which has wide = 1 now cannot hold chars, which needs multibyte encoding.

Author: Oldes

src/core/s-make.c

@@ -319,7 +319,7 @@ x*/	REBCNT Insert_Value(REBSER *series, REBCNT index, REBVAL *item, REBCNT type,
 	if (!BYTE_SIZE(src)) {
 		up = UNI_SKIP(src, index);
 		for (n = 0; n < length; n++)
-			if (up[n] > 0xff) break;
+			if (up[n] >= 0x80) break;
 		if (n < length) wide = sizeof(REBUNI);
 	}
 

src/core/s-unicode.c

@@ -842,7 +842,7 @@ ConversionResult ConvertUTF8toUTF32 (
 		if ((ch = *src) >= 0x80) {
 			ch = Decode_UTF8_Char(&src, &len);
 			if (ch == 0) ch = UNI_REPLACEMENT_CHAR; // temporary!
-			if (ch > 0xff) flag = 1;
+			if (ch >= 0x80) flag = 1;
 		} if (ch == CR && ccr) {
 			if (src[1] == LF) continue;
 			ch = LF;