Mark BytesMut::extend_from_slice as inline

[samanpa]

BytesMut::extend_from_slice , <BytesMut as BufMut>::put_u8 and <BytesMut as BufMut>::put_slice can be bottlenecks when encoding certain messages. This PR just marks extend_from_slice as inline since it is probably an easier change to accept.

example benchmark with these changes applied. Involves encoding a struct in ASN.1

$ hyperfine ./target/release/benchber ./target/release/benchber.orig 
Benchmark 1: ./target/release/benchber
  Time (mean ± σ):      1.179 s ±  0.067 s    [User: 0.937 s, System: 0.226 s]
  Range (min … max):    1.055 s …  1.257 s    10 runs
 
Benchmark 2: ./target/release/benchber.orig
  Time (mean ± σ):      1.935 s ±  0.101 s    [User: 1.686 s, System: 0.233 s]
  Range (min … max):    1.795 s …  2.159 s    10 runs

Summary
  './target/release/benchber' ran
    1.64 ± 0.13 times faster than './target/release/benchber.orig