Improve f64 encoding

Copy the bytes into a buffer instead of destructuring them. On my machine
it's an about 15% performance improvement.

A benchmark was added, which can be run via

    cargo bench --bench f64_enc --features use_alloc

Author: vmx

Cargo.toml

@@ -42,3 +42,9 @@ name = "de"
 path = "benches/bench_de.rs"
 harness = false
 required-features = [ "serde1" ]
+
+[[bench]]
+name = "f64_enc"
+path = "benches/f64_enc.rs"
+harness = false
+required-features = [ "use_alloc" ]

benches/f64_enc.rs

@@ -0,0 +1,17 @@
+use std::hint::black_box;
+
+use cbor4ii::core::{enc::Encode, utils::BufWriter};
+use criterion::{criterion_group, criterion_main, Criterion};
+
+fn bench_f64_enc(c: &mut Criterion) {
+    c.bench_function("f64-enc", |b| {
+        let float: f64 = 12345.6778;
+        b.iter(|| {
+            let mut writer = BufWriter::new(Vec::new());
+            black_box(float.encode(&mut writer).unwrap());
+        })
+    });
+}
+
+criterion_group!(f64_enc, bench_f64_enc);
+criterion_main!(f64_enc);

src/core/enc.rs

@@ -376,8 +376,9 @@ impl Encode for f32 {
 impl Encode for f64 {
     #[inline]
     fn encode<W: Write>(&self, writer: &mut W) -> Result<(), Error<W::Error>> {
-        let [x0, x1, x2, x3, x4, x5, x6, x7] = self.to_be_bytes();
-        writer.push(&[marker::F64, x0, x1, x2, x3, x4, x5, x6, x7])?;
+        let mut buf = [marker::F64, 0, 0, 0, 0, 0, 0, 0, 0];
+        buf[..8].copy_from_slice(&self.to_be_bytes());
+        writer.push(&buf)?;
         Ok(())
     }
 }