Reduce allocations for path conversions on Windows

Previously, UTF-8 to UTF-16 Path conversions on Windows unnecessarily allocate twice, as described in #96297. This commit fixes that issue.

Author: AronParker

library/std/src/sys/windows/mod.rs

@@ -156,7 +156,13 @@ pub fn unrolled_find_u16s(needle: u16, haystack: &[u16]) -> Option<usize> {
 
 pub fn to_u16s<S: AsRef<OsStr>>(s: S) -> crate::io::Result<Vec<u16>> {
     fn inner(s: &OsStr) -> crate::io::Result<Vec<u16>> {
-        let mut maybe_result: Vec<u16> = s.encode_wide().collect();
+        // Most paths are ASCII, so reserve capacity for as much as there are bytes
+        // in the OsStr plus one for the null-terminating character. We are not
+        // wasting bytes here as paths created by this function are primarily used
+        // in an ephemeral fashion.
+        let mut maybe_result: Vec<u16> = Vec::with_capacity(s.len() + 1);
+        maybe_result.extend(s.encode_wide());
+
         if unrolled_find_u16s(0, &maybe_result).is_some() {
             return Err(crate::io::const_io_error!(
                 ErrorKind::InvalidInput,