read_to_end should fit >=MAX_LEN_UTF8 in buf

On Windows, the UTF-16 to UTF-8 translation is made simpler by ensuring we don't split code points.

Author: ChrisDenton

library/std/src/io/mod.rs

@@ -464,7 +464,8 @@ pub(crate) fn default_read_to_end<R: Read + ?Sized>(
             }
         }
 
-        if buf.len() == buf.capacity() {
+        // Ensure there's at least enough space for one UTF-8 encoded code point.
+        if buf.spare_capacity_mut().len() < char::MAX_LEN_UTF8 {
             // buf is full, need more space
             buf.try_reserve(PROBE_SIZE)?;
         }