feat: use slab allocation for strings and bytes

[nrwiersma]

This PR adds support for slab allocation for strings and bytes smaller than 1024 bytes, but will likely be all of them.
This creates a new fast path and centralises the read code, and it is now basically identical.

The benchmarked changes are as follows:

name                      old time/op    new time/op    delta
SuperheroDecode-8            281ns ± 0%     245ns ± 2%   -12.81%  (p=0.000 n=8+10)
SuperheroEncode-8            202ns ± 1%     203ns ± 3%      ~     (p=0.255 n=10+10)
PartialSuperheroDecode-8     119ns ± 1%     111ns ± 1%    -6.82%  (p=0.000 n=10+10)
SuperheroGenericDecode-8    1.44µs ± 1%    1.38µs ± 0%    -3.63%  (p=0.000 n=9+9)
SuperheroGenericEncode-8     258ns ± 1%     278ns ± 1%    +7.84%  (p=0.000 n=9+9)
SuperheroWriteFlush-8        167ns ± 1%     169ns ± 1%    +1.06%  (p=0.023 n=9+10)

name                      old alloc/op   new alloc/op   delta
SuperheroDecode-8            64.0B ± 0%     47.0B ± 0%   -26.56%  (p=0.000 n=10+10)
SuperheroEncode-8             112B ± 0%      112B ± 0%      ~     (all equal)
PartialSuperheroDecode-8     16.0B ± 0%      9.0B ± 0%   -43.75%  (p=0.000 n=10+10)
SuperheroGenericDecode-8    1.70kB ± 0%    1.68kB ± 0%    -1.00%  (p=0.000 n=10+10)
SuperheroGenericEncode-8     80.0B ± 0%     80.0B ± 0%      ~     (all equal)
SuperheroWriteFlush-8        0.00B          0.00B           ~     (all equal)

name                      old allocs/op  new allocs/op  delta
SuperheroDecode-8             4.00 ± 0%      0.00       -100.00%  (p=0.000 n=10+10)
SuperheroEncode-8             1.00 ± 0%      1.00 ± 0%      ~     (all equal)
PartialSuperheroDecode-8      1.00 ± 0%      0.00       -100.00%  (p=0.000 n=10+10)
SuperheroGenericDecode-8      33.0 ± 0%      29.0 ± 0%   -12.12%  (p=0.000 n=10+10)
SuperheroGenericEncode-8      3.00 ± 0%      3.00 ± 0%      ~     (all equal)
SuperheroWriteFlush-8         0.00           0.00           ~     (all equal)

[Ullaakut]

SGTM, I just find it weird to pass a string of the type for error reporting that does not seem to be specific to the returned type.

[nrwiersma]

It allows specific errors, which are helpful. Can be renamed to op perhaps