Let str::replace take a pattern

It appears this was left out of RFC #528 because it might be useful to
also generalize the second argument in some way.  That doesn't seem to
prevent generalizing the first argument now, however.

This is a [breaking-change] because it could cause type-inference to
fail where it previously succeeded.

src/libcollections/str.rs

@@ -1706,7 +1706,7 @@ impl str {
     /// assert_eq!(s, s.replace("cookie monster", "little lamb"));
     /// ```
     #[stable(feature = "rust1", since = "1.0.0")]
-    pub fn replace(&self, from: &str, to: &str) -> String {
+    pub fn replace<'a, P: Pattern<'a>>(&'a self, from: P, to: &str) -> String {
         let mut result = String::new();
         let mut last_end = 0;
         for (start, part) in self.match_indices(from) {

src/libcollectionstest/str.rs

@@ -269,6 +269,15 @@ fn test_replace_2d() {
     assert_eq!(data.replace(d, repl), data);
 }
 
+#[test]
+fn test_replace_pattern() {
+    let data = "abcdαβγδabcdαβγδ";
+    assert_eq!(data.replace("dαβ", "😺😺😺"), "abc😺😺😺γδabc😺😺😺γδ");
+    assert_eq!(data.replace('γ', "😺😺😺"), "abcdαβ😺😺😺δabcdαβ😺😺😺δ");
+    assert_eq!(data.replace(&['a', 'γ'] as &[_], "😺😺😺"), "😺😺😺bcdαβ😺😺😺δ😺😺😺bcdαβ😺😺😺δ");
+    assert_eq!(data.replace(|c| c == 'γ', "😺😺😺"), "abcdαβ😺😺😺δabcdαβ😺😺😺δ");
+}
+
 #[test]
 fn test_slice() {
     assert_eq!("ab", &"abc"[0..2]);