VM: optimizing slice

[petar-dambovaliev]

Optimizing the internal representation of gno slices

A custom data structure where it either holds an array of const N or the current slice implementation.
The empty slots in the array are filled with nil.
When push number N + 1 happens, the elements are moved from the array to the slice.
Statistically, most slices/vectors are of small size. So a reasonable N can be anything up to 10.

This could potentially have several benefits.

1. Appending up to N does not cause reallocation
2. Removed 1 layer of heap indirection
3. Compiler optimizations