[wip] feat: main stf

[spiral-ladder]

This PR contains major changes that overhaul how base branch implements state transition, which mainly is a direct lift of the TS version.

Proposed changes in this PR

No getters/setters in types

Most of what getters/setters accomplish can be accomplished similar with a pointer to the type, eg. *Type. In this PR, I propose having 'getter's of 2 variants. As an example:

    pub fn slot(self: *const BeaconStateAllForks) u64 {
        return switch (self.*) {
            inline else => |state| state.slot,
        };
    }

    pub fn slotPtr(self: *const BeaconStateAllForks) *u64 {
        return switch (self.*) {
            inline else => |state| &state.slot,
        };
    }

We use slot() for immutable access to state.slot, and slotPtr() if we want to change its value. I think realistically, using only 1 function that returns a pointer to the variable (in this case, only supporting slotPtr()) should have negligible overhead, but isn't aesthetic to deref every time we need to access the value. Having a separate *Ptr() function also implies that the value is subject to change.

Simplifying types

In the base branch, we have a separate types.zig that imports some ssz types such as ssz.primitive.Root, ssz.primitive.Epoch and so on, and reexports them. I think this is unnecessary. We can import them directly from the consensus_types module.

Breaking up utils

I'm personally not a fan of naming things utils because they usually could 100% be better named, eg. if the functions in the utils have to do with block processing, just call it block_processing.zig or at the very least block_utils.zig. Otherwise, they just end up as a dumping ground for functions that should belong in other files that we don't know how to name, and we end up with a file/directory with a random assortment of things. Another advantage of not naming things utils is that at first glance we know what code lives in a file by its filename.

[github-actions]

Performance Report

✔️ no performance regression detected

Full benchmark results

Benchmark suite	Current: 9817885	Previous: null	Ratio
get values - 1000	338.00 ns/op
get values - naive - 1000	688.00 ns/op
set values - 1000	322.00 ns/op
set values - naive - 1000	634.00 ns/op
get values - 1000000	849.00 ns/op
get values - naive - 1000000	1.4730 us/op
set values - 1000000	880.00 ns/op
set values - naive - 1000000	2.1510 us/op
JS - computeSyncCommitteeIndices - 16384 indices	320.18 ms/op
Zig - computeSyncCommitteeIndices - 16384 indices	3.9825 ms/op
JS - computeSyncCommitteeIndices - 250000 indices	292.22 ms/op
Zig - computeSyncCommitteeIndices - 250000 indices	16.884 ms/op
JS - computeSyncCommitteeIndices - 1000000 indices	310.14 ms/op
Zig - computeSyncCommitteeIndices - 1000000 indices	32.477 ms/op
JS - unshuffleList - 16384 indices	938.54 us/op
Zig - unshuffleList - 16384 indices	579.08 us/op
JS - unshuffleList - 250000 indices	13.844 ms/op
Zig - unshuffleList - 250000 indices	8.7613 ms/op
JS - unshuffleList - 1000000 indices	54.701 ms/op
Zig - unshuffleList - 1000000 indices	35.154 ms/op

by benchmarkbot/action