refactor(mpz-garble-core): batched garbling

[sinui0]

This PR refactors mpz-garble-core and implements stack-allocated batching. On my machine these changes result in ~50% improved garbling performance.

Changes

• Reduce allocation overhead by reusing the memory between garblings.
• Replace Option<Label> with just Label and use default. This removes a bunch of branching to check whether the value is set.
• Add stack allocated batching of the encrypted gates
• Tweaked the control flow for iterating over the gates