Comparing metrics with benchstat

In continuation of the message processing comparisons yesterday, the benchmarks below performed additional tests.

Based on the benchmark:

1. There is less allocated bytes and instances per operation with preallocated version (B/op and allocs/op)
2. There is less time to process each operation and higher throughput with preallocated version (sec/op and B/s)

go test -run '^$' \
  -bench '^BenchmarkBuildMessageBatch$' \
  -count=10 \
  > /tmp/codec01-message-batch.txt

benchstat \
  -filter '.unit:(B/op OR allocs/op)' \
  -col /implementation \
  -row .name \
  /tmp/codec01-message-batch.txt
goos: linux
goarch: arm64
pkg: codec01
                  │    growing    │             preallocated             │
                  │     B/op      │     B/op      vs base                │
BuildMessageBatch   278.69Ki ± 0%   64.00Ki ± 0%  -77.04% (p=0.000 n=10)

                  │   growing   │            preallocated            │
                  │  allocs/op  │ allocs/op   vs base                │
BuildMessageBatch   19.000 ± 0%   1.000 ± 0%  -94.74% (p=0.000 n=10)

benchstat \
  -filter '.unit:(ns/op OR MB/s)' \
  -col /implementation \
  -row .name \
  /tmp/codec01-message-batch.txt
goos: linux
goarch: arm64
pkg: codec01
                  │   growing    │            preallocated             │
                  │    sec/op    │   sec/op     vs base                │
BuildMessageBatch   62.29µ ± 14%   14.73µ ± 7%  -76.35% (p=0.000 n=10)

                  │    growing    │              preallocated              │
                  │      B/s      │      B/s       vs base                 │
BuildMessageBatch   980.0Mi ± 12%   4144.2Mi ± 7%  +322.87% (p=0.000 n=10)