Refactor get_* calls in harnesses for fewer steps

[dkcumming]

This PR refactors calls to get_{account, mint, multisig} where possible in all proof harnesses to reduce the amount of steps proofs need to take. A selection of proofs were run again to see the reduction in steps, not every harness could be refactored and some proofs go for long without stopping so it is hard to see how the count changes.

The naming convention introduced is inspired by specification languages that have an old(...) operator to talk about pre-strate (e.g. Dafny). In the post state I added new to variable so it would be clear this is captured after the implementation has executed.

Proofs run and change in steps:

• initialize_account: 89207 -> 89207 = no change, probably indicates not getting into the else branch
• transfer: 4011 -> 3459 = 14% reduction
• mint_to: 3248 -> 2834 = 13% reduction
• burn: 4699 -> 3457 = 26% reduction
• close_account: 3711 -> 3021 = 19% reduction
• transfer_checked: 4755 -> 4065 = 15% reduction
• burn_checked: 5290 -> 3910 = 26% reduction
• approve: 3049 -> 2773 = 9% reduction
• approve_checked: 3793 -> 3517 = 7% reduction
• revoke: 2409 -> 2133 = 11% reduction
• freeze_account: 3199 -> 2647 = 17% reduction
• thaw_account: 3199 -> 2647 = 17% reduction
• mint_to_checked: 3689 -> 3275 = 11% reduction
• sync_native: 2331 -> 2055 = 12% reduction

The concrete test harnesses that are able to run with our restricted harnesses were also run successfully

[dkcumming]

Concrete tests have been run post merge for both p-token and programs and passed all (except the malformed input tests we do not handle)

[jberthold]

LGTM
Nice refactoring, code is much easier to read/follow now.