[WIP] Herbert/use side channel for state changes

[HerbertJordan]

The idea of this PR is to remove the hardly modified state from the data that is propagated copied and checked in the interpreter loop. Instead, the program pointer (pc) is used to signal termination.

On Clang: 14.0.0, based on #161

goos: linux
goarch: amd64
pkg: github.com/Fantom-foundation/Tosca/go/vm/test
cpu: Intel(R) Core(TM) i7-6600U CPU @ 2.60GHz
                            │ track_size_explicit_clang.log │     state_on_the_side_clang.log     │
                            │            sec/op             │    sec/op     vs base               │
Inc/1/evmzero-4                                1.819µ ±  8%   1.962µ ± 36%        ~ (p=0.097 n=7)
Inc/10/evmzero-4                               1.831µ ±  7%   1.771µ ±  6%        ~ (p=0.057 n=7)
Fib/1/evmzero-4                                1.748µ ±  5%   1.680µ ±  3%   -3.89% (p=0.005 n=7)
Fib/5/evmzero-4                                5.396µ ±  7%   5.102µ ±  7%        ~ (p=0.073 n=7)
Fib/10/evmzero-4                               43.90µ ± 12%   41.48µ ±  8%   -5.51% (p=0.026 n=7)
Fib/15/evmzero-4                               511.4µ ±  8%   456.6µ ± 12%  -10.72% (p=0.038 n=7)
Fib/20/evmzero-4                               4.721m ±  9%   4.629m ±  6%        ~ (p=0.097 n=7)
Sha3/1/evmzero-4                               1.402µ ± 10%   1.404µ ±  5%        ~ (p=0.710 n=7)
Sha3/10/evmzero-4                              2.220µ ±  6%   2.177µ ±  4%        ~ (p=0.259 n=7)
Sha3/100/evmzero-4                             10.50µ ± 14%   10.16µ ±  5%        ~ (p=0.120 n=7)
Sha3/1000/evmzero-4                            91.32µ ± 11%   89.04µ ±  5%        ~ (p=0.259 n=7)
Arith/1/evmzero-4                              1.996µ ±  4%   1.964µ ±  6%        ~ (p=0.103 n=7)
Arith/10/evmzero-4                             4.194µ ±  8%   4.034µ ±  5%   -3.81% (p=0.017 n=7)
Arith/100/evmzero-4                            25.47µ ±  9%   24.20µ ±  6%        ~ (p=0.165 n=7)
Arith/280/evmzero-4                            70.09µ ±  7%   66.58µ ±  9%        ~ (p=0.209 n=7)
Memory/1/evmzero-4                             2.422µ ±  5%   2.341µ ±  4%   -3.34% (p=0.026 n=7)
Memory/10/evmzero-4                            4.579µ ±  8%   4.378µ ±  7%   -4.39% (p=0.017 n=7)
Memory/100/evmzero-4                           25.18µ ±  6%   25.07µ ±  8%        ~ (p=0.383 n=7)
Memory/1000/evmzero-4                          258.2µ ± 10%   235.8µ ± 10%        ~ (p=0.053 n=7)
Memory/10000/evmzero-4                         2.240m ± 33%   2.091m ±  7%   -6.66% (p=0.017 n=7)
Analysis/jumpdest/evmzero-4                    1.263µ ±  9%   1.262µ ±  6%        ~ (p=0.712 n=7)
Analysis/stop/evmzero-4                        1.262µ ±  7%   1.276µ ±  4%        ~ (p=0.804 n=7)
Analysis/push1/evmzero-4                       1.255µ ±  4%   1.263µ ±  5%        ~ (p=0.514 n=7)
Analysis/push32/evmzero-4                      1.307µ ±  5%   1.269µ ±  7%        ~ (p=0.535 n=7)
geomean                                        11.88µ         11.52µ         -3.01%

[LuisPH3]

Why a pointer and not a reference? if you choose pointer, there must be a good reason I want to know.