Is your feature request related to a problem or challenge? Please describe what you are trying to do.

We can precalculate the capacity of the value in take_bytes based on the offsets, while building the offsets.
This is faster than letting the buffer allocate for us.
As a bonus, we also save some memory by precisely allocating the resulting buffer, instead of overallocating.

Describe the solution you'd like

Describe alternatives you've considered

Additional context