Run integration tests against a live node

[digorithm]

Problem

Sometimes, we catch integration errors with fuel-core only when users try using the SDK against a live node (e.g., beta-4), which causes frustration and takes us more time than it should to fix these issues. These issues aren't found when testing against a local node, even if the node's version is at the latest, because the live node might have been configured differently.

Potential solution

We need to catch these errors as early as possible by running some of our integration tests (tests that interact with a provider) against the live node and not only against a local version of the node. However, running these integration tests against a live node would introduce potential flakiness and higher latency due to all the network hops inside our CI runs. Doing this inside every pull request could slow us down unnecessarily.

An approach to this could be to run these tests against a live node only when a PR is merged to master; the CI would trigger a run against the new master branch and run these tests against a live node. This way, it wouldn't block or slow down a PR, and since the master branch hasn't been released yet, catching an issue that's happening only against a live node at this stage isn't problematic, and we can start working on it before this issue is caught after a release.