Make TokenStream::from_iter less general and more efficient.

The current code has this impl:
```
impl<T: Into<TokenStream>> iter::FromIterator<T> for TokenStream
```
If given an `IntoIterator<Item = TokenTree>`, it will convert each individual
`TokenTree` to a `TokenStream` (at the cost of two allocations: a `Vec`
and an `Lrc`). It will then merge those `TokenStream`s into a single
`TokenStream`. This is inefficient.

This commit changes the impl to this less general one:
```
impl iter::FromIterator<TokenTree> for TokenStream
```
It collects the `TokenTree`s into a single `Vec` first and then converts that
to a `TokenStream` by wrapping it in a single `Lrc`. The previous generality
was unnecessary; no other code needs changing.

This change speeds up several benchmarks by up to 4%.

Author: nnethercote

src/libsyntax/tokenstream.rs

@@ -202,9 +202,9 @@ impl From<TokenTree> for TreeAndJoint {
     }
 }
 
-impl<T: Into<TokenStream>> iter::FromIterator<T> for TokenStream {
-    fn from_iter<I: IntoIterator<Item = T>>(iter: I) -> Self {
-        TokenStream::from_streams(iter.into_iter().map(Into::into).collect::<SmallVec<_>>())
+impl iter::FromIterator<TokenTree> for TokenStream {
+    fn from_iter<I: IntoIterator<Item = TokenTree>>(iter: I) -> Self {
+        TokenStream::new(iter.into_iter().map(Into::into).collect::<Vec<TreeAndJoint>>())
     }
 }