feat: use vtpool wherever possible

[kruskall]

Closes #130

[kruskall]

I've run the benchmark and compared with main:

                            │   old.txt   │               new.txt               │
                            │   sec/op    │   sec/op     vs base                │
MapToModelPb/error-model-20   5.093µ ± 2%   2.089µ ± 2%  -58.99% (p=0.000 n=10)

                            │   old.txt   │              new.txt               │
                            │    B/op     │    B/op     vs base                │
MapToModelPb/error-model-20   5681.5 ± 0%   929.0 ± 0%  -83.65% (p=0.000 n=10)

                            │  old.txt   │              new.txt               │
                            │ allocs/op  │ allocs/op   vs base                │
MapToModelPb/error-model-20   75.00 ± 0%   26.00 ± 0%  -65.33% (p=0.000 n=10)

[kruskall]

I've run the benchmark and compared with main:

NOTE: both benchmarks are reusing an event by calling ResetVT so in theory the actual performance of main is slower than what's reported above.

[kruskall]

This should be completely merged in now 🎉