ci: Debug CLI / test-solana job

[nvsriram]

CLI / test-solana job has been failing sporadically (race condition probably) in recent PRs so this PR attempts to fix it

Context:
In the Solana CLI test, after upgrading to 2.0.0 from 1.0.0, the upgrade command calls nttFromManager which in turn calls getProtocol to fetch the program and version (via getVersion). The default getVersion output on error/ fetching a program not deployed is 3.0.0 (as that is the current version on main). This causes a diff between the actual deployed version and fetched version that ultimately results in the failure.

Fix:

1. ❌ Revert getVersion default to be 2.0.0. This is not ideal as it does not reflect the current version on main.
2. ✅ Add a sleep/spinloop after we deploy the program and before we fetch the program to ensure we fetch the version directly from the deployed program rather than depending on the default value returned.

[evan-gray]

Is there a less arbitrary way to detect this? Can it be known that the program will always take less than 2 seconds to be deployed?

[nvsriram]

There is no guarantee that the program will always take less than 2 seconds to be deployed or in this case for the RPC to catch up to the deployed program (I could probably update the comment).
This sleep happens after deploy proc has exited. So this sleep just acts as a buffer for the RPC to query the correct deployed program - similar to the sleep before initialize.

The less arbitrary option I considered was a spinloop inside nttFromManager if an optional expectedVersion parameter is passed in:

async function nttFromManager<N extends Network, C extends Chain>(
    ch: ChainContext<N, C>,
    nativeManagerAddress: string,
    expectedVersion?: string
): Promise<{ ntt: Ntt<N, C>; addresses: Partial<Ntt.Contracts> }> {
    let onlyManager: Ntt<N, C>;
    onlyManager = await ch.getProtocol("Ntt", {
        ...
    });
    while (expectedVersion && getVersion(ch.chain, onlyManager) !== expectedVersion) {
       await new Promise((resolve) => setTimeout(resolve, 2000));
       onlyManager = await ch.getProtocol("Ntt", {
         ...
       });
    }
    ...
}

And it could be called in upgrade:

const { ntt: upgraded } = await nttFromManager(ch, chainConfig.manager, toVersion ?? undefined);

Concerns with this approach:

1. This CI job fail is only due to the edge case where the CLI release has not caught up to the latest NTT manager version so the complexity added with a spinloop and debugging might be overkill?
2. Calling getProtocol in a loop might not be the most optimal way to fetch versions
3. Not sure if the strict equality is the best way to compare versions (in case of semver etc.)

[evan-gray]

I think my proposal in #574 fixes the underlying problem without the arbitrary wait.

[evan-gray]

closing in favor of #574