Reduce mpc contract latency

[kevindeforth]

Resolves #131 :

• add contract binaries to compiled-contracts
• add pytest fixtures for contract compilation
• improve pytests for contract update
• add a pytest for state cleanup
• changes the gas requirements for sign call from 50 teragas to 15

Benchmarks:
For 40 signatures:

Contract	avg. receipts	avg. gas [Tgas]
V0	8	11.30479597562405
V1	4	6.131075775468398

Contract Logic changes:
Each row is a promise, ignore empty rows.

V0	V1
do some checks	do some checks, add request to state, schedule yield promise
add request to state, schedule yield promise
remove signature request from state	clean state & return signature or error
return signature or error

Additional Contract State:
Because of the new logic, failed signature requests need to be periodically removed from the state.
This requires additional contract state:

    request_by_timestamp: Vector<(u64, SignatureRequest)>,
    config: ConfigV1,

request_by_timestamp is iterable and stores the env::block_height() during which the request was added to the state.
config contains is the timeout after which signature requests can be removed from the contract, it also defines the maximum number of requests that can be removed during a call to sign(), respond() or remove_timed_out_requests().

Note:
The signature of the sign function changes with this PR, specifically, the return value is no longer Result<near_sdk::Promise, Error>, but all Errors result in panic.

edit:
CI time:
The new pytest takes a long time and increases the ci time to 40 minutes, which seems quite long. This is expected to increase as we add more tests. If reviewers agree, I would like to restrict the ci to pytest -m "not slow", skipping all tests that are marked as slow via @pytest.mark.slow
Alternatively, we could add a marker ci to annotate tests that should be included in ci, ignoring all tests that do not have that marker.

[bowenwang1996]

Meta comment: given that V1 is not live yet, let's not introduce V2 unnecessarily and collapse changes in V2 into V1

[saketh-are]

Should this be removed now that sign takes the responsibility for cleanup?

[saketh-are]

In my previous testing I was not able to produce nonce conflicts in localnet even under significant load (10s or 100s of txs). I was running the other pytest test_signature_request with high num_requests and didn't see any requests fail to be included. I believe this is because the local setup is too simple for transaction reordering to happen. So I wonder if there is some other underlying cause for flakiness observed here.

[kevindeforth]

In the timespan between the rpc request to fetch the current nonce value and until the update transaction is sent, the mpc participant already submitted other transactions and the nonce increased.

This improved significantly after using release version of near core, but I still get the error every now an then.

Still learning about keys and accounts. Does each access key have its own nonce and would it help to use a different access key for the update calls?

[kevindeforth]

I opened an issue for this, as this PR is already quite large: #182

[saketh-are]

Thanks @kevindeforth!