perf: switch to json decoder and reduce heap usage

[kruskall]

This PR aims to solve the recent increase in heap usage without sacrificing some of the performance improvements we gained.

In details:

• we are flattening the map and storing the bulk response item in a slice since we are not using the map keys.
• reduce allocations over the reflection-based approach of the std lib
• optimize correct/common case by only adding 1 to the counter of indexed request. We use the slice only for failed requests which should ideally be 0 or really low compared to the total
• do not use io.ReadAll and switch back to a decoder to avoid storing temporary the whole response in memory

Benchstat output compared to the current implementation:

                                      │  before.txt  │              after.txt               │
                                      │    sec/op    │    sec/op     vs base                │
Appender/NoCompression-20               479.5n ±  5%   494.4n ±  4%        ~ (p=0.190 n=10)
Appender/NoCompressionScaling-20        868.8n ± 10%   838.3n ± 42%        ~ (p=0.912 n=10)
Appender/BestSpeed-20                   824.3n ±  4%   673.6n ±  2%  -18.29% (p=0.000 n=10)
Appender/BestSpeedScaling-20            692.9n ±  9%   673.0n ±  9%   -2.88% (p=0.011 n=10)
Appender/DefaultCompression-20          1.379µ ±  5%   1.165µ ±  4%  -15.52% (p=0.000 n=10)
Appender/DefaultCompressionScaling-20   1.414µ ±  3%   1.175µ ±  3%  -16.90% (p=0.000 n=10)
Appender/BestCompression-20             1.438µ ±  5%   1.177µ ±  4%  -18.12% (p=0.000 n=10)
Appender/BestCompressionScaling-20      1.395µ ±  3%   1.167µ ±  2%  -16.31% (p=0.000 n=10)
geomean                                 990.9n         878.0n        -11.40%

                                      │  before.txt  │               after.txt                │
                                      │     B/s      │      B/s        vs base                │
Appender/NoCompression-20               266.5Mi ± 5%    258.5Mi ±  4%        ~ (p=0.190 n=10)
Appender/NoCompressionScaling-20        147.1Mi ± 9%    152.4Mi ± 73%        ~ (p=0.912 n=10)
Appender/BestSpeed-20                   155.0Mi ± 4%    189.7Mi ±  2%  +22.39% (p=0.000 n=10)
Appender/BestSpeedScaling-20            184.4Mi ± 8%    189.9Mi ± 10%   +2.97% (p=0.011 n=10)
Appender/DefaultCompression-20          92.70Mi ± 5%   109.77Mi ±  4%  +18.41% (p=0.000 n=10)
Appender/DefaultCompressionScaling-20   90.40Mi ± 4%   108.77Mi ±  3%  +20.32% (p=0.000 n=10)
Appender/BestCompression-20             88.88Mi ± 6%   108.60Mi ±  4%  +22.18% (p=0.000 n=10)
Appender/BestCompressionScaling-20      91.65Mi ± 3%   109.51Mi ±  2%  +19.48% (p=0.000 n=10)
geomean                                 129.0Mi         145.6Mi        +12.87%

                                      │ before.txt │              after.txt              │
                                      │    B/op    │    B/op      vs base                │
Appender/NoCompression-20               333.0 ± 0%   202.0 ±  0%  -39.34% (p=0.000 n=10)
Appender/NoCompressionScaling-20        340.5 ± 2%   210.5 ±  5%  -38.18% (p=0.000 n=10)
Appender/BestSpeed-20                   708.0 ± 2%   214.0 ±  0%  -69.77% (p=0.000 n=10)
Appender/BestSpeedScaling-20            692.5 ± 2%   216.5 ±  2%  -68.74% (p=0.000 n=10)
Appender/DefaultCompression-20          676.0 ± 4%   206.0 ±  9%  -69.53% (p=0.000 n=10)
Appender/DefaultCompressionScaling-20   673.5 ± 5%   207.0 ± 11%  -69.27% (p=0.000 n=10)
Appender/BestCompression-20             676.0 ± 5%   204.5 ±  7%  -69.75% (p=0.000 n=10)
Appender/BestCompressionScaling-20      673.0 ± 5%   208.5 ±  9%  -69.02% (p=0.000 n=10)
geomean                                 572.3        208.6        -63.56%

                                      │ before.txt │              after.txt              │
                                      │ allocs/op  │ allocs/op   vs base                 │
Appender/NoCompression-20               2.000 ± 0%   4.000 ± 0%  +100.00% (p=0.000 n=10)
Appender/NoCompressionScaling-20        2.000 ± 0%   4.000 ± 0%  +100.00% (p=0.000 n=10)
Appender/BestSpeed-20                   2.000 ± 0%   4.000 ± 0%  +100.00% (p=0.000 n=10)
Appender/BestSpeedScaling-20            2.000 ± 0%   4.000 ± 0%  +100.00% (p=0.000 n=10)
Appender/DefaultCompression-20          2.000 ± 0%   4.000 ± 0%  +100.00% (p=0.000 n=10)
Appender/DefaultCompressionScaling-20   2.000 ± 0%   4.000 ± 0%  +100.00% (p=0.000 n=10)
Appender/BestCompression-20             2.000 ± 0%   4.000 ± 0%  +100.00% (p=0.000 n=10)
Appender/BestCompressionScaling-20      2.000 ± 0%   4.000 ± 0%  +100.00% (p=0.000 n=10)
geomean                                 2.000        4.000       +100.00%