Store merkle positions in binary rather than using RKYV serialisation

[miloszm]

Summary

Storing as RKYV serialisation does not scale well for large files, as it requires extra cloning.
It will be more efficient to use binary representation.
Also, some optimisations can be made like storing merkle contract id as u32 rather than u64 as its upper 32 bits are always zero.

Possible solution design or implementation

Implement both and compare results. If time savings are substantial, switch to the new method.
Maintain the old method for backwards compatibility, so if the new file does not exist, use the old one.