CI job template_regression_tests could be faster

[blmalone]

We do chain discovery via the address registry for each task that is simulated. If we could either (1) make chain discovery faster with concurrency or (2) read the addresses from the local addresses.json file (down side is we are running a no production code path).

See PR that introduced this feature: https://github.com/ethereum-optimism/superchain-ops/pull/634