[Refactor] simplify and unify the storage setup for tests

[ljedrz]

The hybrid nature of our storage makes testing it tricky, and there are several spots where we are currently using a mix of in-memory and persistent storage, which complicates things further. The changes proposed in this PR are something that I've hoped to see in snarkVM for a while now, and since some other, concurrent snarkVM work requires a stricter separation between the storage variants, it was a good opportunity to finally propose them.

The following is a rough summary of these changes:

• all the tests now implicitly either use the persistent storage or the in-memory one; there is no mixing, unless it's done explicitly
• as far as storage is concerned, all the tests can now run concurrently
• the open_(.*)testing methods are no longer needed, there is no test special-casing for rocksdb either
• the open methods no longer require S to be Clone
• instead of a static DB, there is now a static DATABASES, which has more than a single member only in tests
• the ledger-with-rocksdb CI job should be runnable again

Future improvements:

• if the open methods for the stores didn't rely so much on StorageMode (i.e. if we could use an instance of actual storage to open the stores), we wouldn't need static DATABASES (or DB, as is the case currently); that being said, I'd prefer it to be done separately

Requires: ProvableHQ/aleo-std#12

[ljedrz]

While that one program (rand_chacha_check) in test_vm_execute_and_finalize is returning a different outcome than the expectation regardless of features (rocksdb, test), I'm not sure why it happens just yet, but the PR may already be reviewed, and hopefully the cause will be revealed in the process.

For reference, this is the one mismatch:

============================================================
TEST
------------------------------------------------------------
program: test_rand.aleo
function: rand_chacha_check
inputs:
- 1field
- true

============================================================
EXPECTED
------------------------------------------------------------
verified: true
execute:
  test_rand.aleo/rand_chacha_check:
    outputs:
    - '{"type":"future","id":"818878742790741579153893179075772445872751227433677932822653185952935999557field","value":"{\n  program_id: test_rand.aleo,\n  function_name: rand_chacha_check,\n  arguments: [\n    1field,\n    true\n  ]\n}"}'
speculate: the execution was rejected
add_next_block: succeeded.

============================================================
ACTUAL
------------------------------------------------------------
verified: true
execute:
  test_rand.aleo/rand_chacha_check:
    outputs:
    - '{"type":"future","id":"818878742790741579153893179075772445872751227433677932822653185952935999557field","value":"{\n  program_id: test_rand.aleo,\n  function_name: rand_chacha_check,\n  arguments: [\n    1field,\n    true\n  ]\n}"}'
speculate: the execution was accepted
add_next_block: succeeded.

============================================================

[ljedrz]

I have a branch ready with a StorageMode::Test proposal, which would reduce the reliance of this PR on the test feature. I'd be happy to update this PR based on it if we can release an update to aleo-std soon.

[ljedrz]

Update: I no longer have that one test mismatch with updated aleo-std, so I'll double-check the other tests and update this PR shortly.

[ljedrz]

I updated the PR to use a new aleo-std-storage feature which makes things simpler and more explicit; it hasn't been merged yet, so there are temporarily some dependencies on a specific branch, but other than that this PR is ready for review now.