Optimze append

[Fabianexe]

The append function is slow compared to the get and set method. The reason for this is that unnecessary copies of the data are done in the process. Namely, the get copy the data and the list append a second time. This leads to high allocation numbers and slow runtime overall.
In this PR the number of copies in the process is minimized without touching the BytesQueue logic. This reduces the allocs and halves the runtime of append without changing on the interface.
Some benchmarks result:

name                          old time/op    new time/op    delta
AppendToCache/1-shards-12       8.37µs ±26%    4.06µs ±23%  -51.48%  (p=0.000 n=99+98)
AppendToCache/512-shards-12     3.23µs ± 6%    1.87µs ± 3%  -42.00%  (p=0.000 n=96+92)
AppendToCache/1024-shards-12    3.49µs ± 4%    2.06µs ± 4%  -40.86%  (p=0.000 n=100+93)
AppendToCache/8192-shards-12    4.04µs ± 3%    2.85µs ± 3%  -29.42%  (p=0.000 n=96+97)

name                          old alloc/op   new alloc/op   delta
AppendToCache/1-shards-12       28.1kB ± 0%    14.1kB ± 0%  -49.84%  (p=0.000 n=100+100)
AppendToCache/512-shards-12     28.2kB ± 1%    14.1kB ± 0%  -49.92%  (p=0.000 n=93+95)
AppendToCache/1024-shards-12    29.4kB ± 4%    14.4kB ± 4%  -50.96%  (p=0.000 n=100+94)
AppendToCache/8192-shards-12    34.0kB ± 2%    19.4kB ± 3%  -43.07%  (p=0.000 n=99+98)

name                          old allocs/op  new allocs/op  delta
AppendToCache/1-shards-12         21.0 ± 0%       2.0 ± 0%  -90.48%  (p=0.000 n=100+100)
AppendToCache/512-shards-12       22.0 ± 0%       3.0 ± 0%  -86.36%  (p=0.000 n=100+100)
AppendToCache/1024-shards-12      22.0 ± 0%       3.0 ± 0%  -86.36%  (p=0.000 n=100+100)
AppendToCache/8192-shards-12      22.0 ± 0%       3.0 ± 0%  -86.36%  (p=0.000 n=100+100)

[siennathesane]

Nice change! Happy to it, thanks!