[DX Streamline] LOTUS_RUN_EXPENSIVE_TESTS periodically outside of PR workflow

[BigLep]

Done Criteria

1. LOTUS_RUN_EXPENSIVE_TESTS are run periodically (e.g., nightly) against master automatically. They aren't run as part of the normal PR flow because of how long they take.
2. There is notification via github issue or Slack if there are any failures so that maintainers can try and identify the culprit change.
3. There is a self-service URL to find the previous results of LOTUS_RUN_EXPENSIVE_TESTS.

Why Important

There are "expensive" tests that test key functionality. This is getting exasperated by NI-PoRep tests which can take 10+ minutes on CI. We want to avoid compromising on test quality because of long CI runs. We want the test coverage but not to be slowed down by it for every CI run.

User/Customer

Contributors and maintainers

Notes

1. This was discussed in #fil-lotus-dev on 2024-06-20 in https://filecoinproject.slack.com/archives/CP50PPW2X/p1718859677050149

[rvagg]

Currently LOTUS_RUN_EXPENSIVE_TESTS=1 is the same as CI=true, which is why I proposed that LOTUS_RUN_VERY_EXPENSIVE_TESTS introduces a new level that we would run periodically.

[rvagg]

master now includes one test that's behind LOTUS_RUN_VERY_EXPENSIVE_TESTS, so this has become a slightly higher priority now

[galargh]

I think it'd be good if there was also a way to request the run of very expensive tests on a PR (e.g. through a label) and if they run when the test itself was modified