Basic manual e2e regression test suite

[mitjat]

This PR adds a script that lets us check the HTTP output of various endpoints against old versions of the indexer.

Alluded to in #273, but this script is NOT integrated into CI. Its major caveat is that it's most informative when you run the indexer against prod data for a good number of rounds, and then check what it outputs. I don't think the script is useful enough to do it on every CI run -- it takes time, and it increases brittleness (because the test needs an external, synced oasis-node).

Changes in the PR:

• tests/e2e_regression/run.sh - the main script
• tests/e2e_regression/expected/* - files with the output as produced currently. I'm not checking these in because of the one-off nature of the script's intended usage.
• analyzer/metadata_registry.go - a patch that lets one set up an analyzer with deterministic outputs: Use a config that runs the consensus+emerald analyzer only up to a certain height (this was already possible before this PR), and use the metadata_registry analyzer against a static intended-for-test branch of the registry.
  • EDIT: I'm excluding this too, now. The static, intended-for-test branch cannot be parsed by the current analyzer. Nobody is sure why (even Jernej, who wrote the tool a good while ago), and I don't think it's important enough to research or fix further. Since expected outputs are not checked in, they don't need to be long-term stable. I added a note about all of this to the main script.

[pro-wh]

alright, but this might rot

[mitjat]

alright, but this might rot

Agreed. I'd say it's mostly valuable for refactoring PRs. I so happened to be working on these more, and found this test useful (and it found a bug!) several times already, and we're likely to do some more refactoring to accommodate multiple runtimes. So I think it has short-term potential :)