Arced Program

[ljedrz]

In one of my ledger-load-only heap profiles the cloning of the Program (ProgramCore<N, Instruction<N>, Command<N>>) was quite prominent, meaning it could be a good candidate for Arcing; such a change would also make the Boxing of the related Deployment superfluous, which would also contribute to a reduction in clones.

The clones of the Program are apparently also quite slow; even in a ledger where this change only results in a tiny change to the number of allocations, this speeds up its loading (via Ledger::load) by ~8%.

I'm filing this PR as a draft for 2 reasons:

• I'm not too happy about the Map-related type changes; while those would be an additional win for in-memory storage, I'm not convinced about the persistent one
• I'd like to perform profiling again with a larger storage, in order to double-check these findings; I will do so soon, when the current devnet's ledger grows larger