test: RPC provider API key modifications

[rvanasa]

Follow-up testing and bugfixes for #252.

This PR...

• Adds state machine tests for the following scenarios:
  • Set the key for a provider that uses RpcAuth::BearerToken.
  • Update a previously existing API key of some provider to None removes it.
  • Panic if trying to specify an API key for a provider that uses RpcAccess::Unauthenticated
  • Panic if a provider is not found.
• Sets up the following canister upgrade tests:
  • Keep previously inserted API keys.
  • Change or keep demo status based on upgrade args.
  • Change or keep principals allowed to manage API key providers based on upgrade args.
• Includes unit tests for BoolStorable and PrincipalStorable.
• Removes unused StringStorable.
• Changes the requestCost canister method to return 0 when in demo mode.
• Fixes the following bugs:
  • BoolStorable had an inverted condition in from_bytes
  • getProviderCost() returned Nat instead of Result<Nat, RpcError> in the state machine test logic

Note that the canister upgrade tests use the same Wasm before and after upgrade. In the future (probably out of scope of this PR), we could extend this to test upgrading from the most recently deployed canister's Wasm file.

[gregorydemay]

Rather than adding those tests regarding implementation of Storable, wouldn't it be better to just update to the latest version of stable structure and remove those wrappers?

[rvanasa]

I timeboxed trying to use the latest version of ic-stable-structures, and the issue that I ran into was that all of the latest versions which implement Storable for Principal and bool use Candid 10.x, which conflicts with transitive dependencies from the forked cketh-minter branch.

After spending quite a lot of time trying to work around this (and running into more and more version conflicts), it might make sense to wait until after we can drop the cketh-minter dependency before making this change in another PR. However, I was able to remove StringStorable to simplify the codebase.

[gregorydemay]

use Candid 10.x, which conflicts with transitive dependencies from the forked cketh-minter branch.

ha I see, another good reason to drop that dependency 😅.

it might make sense to wait until after we can drop the cketh-minter dependency before making this change in another PR

Sure that sounds ok to me. I'm trying to also push on that front, since it will be needed when we want to implement something new (e.g. 3-out-of-4 strategy)

[rvanasa]

Merging and will make any follow-up changes in additional PRs.