Improve the performance of put_u8 and put_i8

[YHM404]

When using put_u8/put_i8, it internally wraps the u8/i8 into &[u8], and then calls put_slice(). Wrapping it into &[u8] is an unnecessary operation. So I optimized this step.

Here's the benchmark below：

Before
test put_u8_bytes_mut ... bench: 333 ns/iter (+/- 70) = 384 MB/s

After
test put_u8_bytes_mut ... bench: 271 ns/iter (+/- 8) = 472 MB/s

[seanmonstar]

This skips some of the length checks making sure there is room for the byte. If the Buf is full, this would be UB.

[YHM404]

This skips some of the length checks making sure there is room for the byte. If the Buf is full, this would be UB.

thx. it's fixed.