Optimize output vector adapter write

[romainreignier]

The first commit of this pull request adds a benchmark for the CBOR serialization in order to compare the change introduced by the second commit of the pull request.

In the output_vector_adapter class, replace the usage of std::copy + std::back_inserter by the method std::vector::insert.
The std::back_inserter generates a lot of calls to std::vector::push_back which allocate the memory on the fly.
For big datasets, usually binary data, the overhead is important.
Resizing the vector in a first step and then copying the data helps a lot.
But from my benchmarks, it appears that using the method .insert() on a vector is almost as performant and shorter to write.

I have added a benchmark similar to the .dump() for JSON but using json::to_cbor() but the performance change is not significant.
But because the application in which I have noticed a bottleneck in the use of std::vector::push_back() was serializaing a lot of binary data to cbor, I have added a benchmark serializing vectors of bytes. For this case, the change proposed is significant.

Original:

2022-07-04T14:43:24+02:00
Running ./json_benchmarks
Run on (8 X 3603.48 MHz CPU s)
CPU Caches:
  L1 Data 32 KiB (x4)
  L1 Instruction 32 KiB (x4)
  L2 Unified 256 KiB (x4)
  L3 Unified 6144 KiB (x1)
Load Average: 0.43, 0.72, 0.85
-----------------------------------------------------------------------------
Benchmark                   Time             CPU   Iterations UserCounters...
-----------------------------------------------------------------------------
BinaryToCbor/8            239 ns          239 ns      2949279 bytes_per_second=107.695M/s
BinaryToCbor/16           264 ns          264 ns      2701934 bytes_per_second=126.646M/s
BinaryToCbor/32           289 ns          289 ns      2419848 bytes_per_second=171.441M/s
BinaryToCbor/64           368 ns          368 ns      1915804 bytes_per_second=217.621M/s
BinaryToCbor/128          502 ns          502 ns      1000000 bytes_per_second=281.356M/s
BinaryToCbor/256          773 ns          773 ns       901424 bytes_per_second=341.569M/s
BinaryToCbor/512         1285 ns         1285 ns       546804 bytes_per_second=395.719M/s
BinaryToCbor/1024        2358 ns         2358 ns       302891 bytes_per_second=422.628M/s
BinaryToCbor/2048        4328 ns         4328 ns       162237 bytes_per_second=455.93M/s
BinaryToCbor/4096        8252 ns         8252 ns        84487 bytes_per_second=475.822M/s
BinaryToCbor/8192       15994 ns        15993 ns        43813 bytes_per_second=489.738M/s
BinaryToCbor/16384      31647 ns        31645 ns        22234 bytes_per_second=494.39M/s
BinaryToCbor/32768      63097 ns        63095 ns        11096 bytes_per_second=495.604M/s

New version with std::vector::insert():

2022-07-04T14:42:37+02:00
Running ./json_benchmarks
Run on (8 X 3786.38 MHz CPU s)
CPU Caches:
  L1 Data 32 KiB (x4)
  L1 Instruction 32 KiB (x4)
  L2 Unified 256 KiB (x4)
  L3 Unified 6144 KiB (x1)
Load Average: 0.34, 0.76, 0.86
-----------------------------------------------------------------------------
Benchmark                   Time             CPU   Iterations UserCounters...
-----------------------------------------------------------------------------
BinaryToCbor/8            232 ns          232 ns      3067496 bytes_per_second=111.04M/s
BinaryToCbor/16           229 ns          229 ns      3011982 bytes_per_second=145.559M/s
BinaryToCbor/32           230 ns          230 ns      3022206 bytes_per_second=215.89M/s
BinaryToCbor/64           231 ns          231 ns      3037105 bytes_per_second=346.82M/s
BinaryToCbor/128          235 ns          235 ns      2979490 bytes_per_second=601.678M/s
BinaryToCbor/256          236 ns          236 ns      2935424 bytes_per_second=1118.63M/s
BinaryToCbor/512          259 ns          259 ns      2712207 bytes_per_second=1.91596G/s
BinaryToCbor/1024         285 ns          285 ns      2489110 bytes_per_second=3.41833G/s
BinaryToCbor/2048         295 ns          295 ns      2364411 bytes_per_second=6.53373G/s
BinaryToCbor/4096         366 ns          366 ns      1925247 bytes_per_second=10.4859G/s
BinaryToCbor/8192         531 ns          531 ns      1319473 bytes_per_second=14.3938G/s
BinaryToCbor/16384       1042 ns         1042 ns       672042 bytes_per_second=14.6641G/s
BinaryToCbor/32768       1853 ns         1853 ns       380673 bytes_per_second=16.4827G/s

If you prefer an implementation with std::vector::resize() + std::memcpy() or std::copy(), the performance is equivalent.

[coveralls]

Coverage remained the same at 100.0% when pulling bb1abe9 on romainreignier:optimize_output_vector_adapter_write into 954b10a on nlohmann:develop.

[nlohmann]

Looks good to me.

[nlohmann]

Thanks!