op-supervisor: initialize cross-safe starting point

[protolambda]

Description

Experimental change to make the op-supervisor require a supervisor_initializeCrossSafe.
This way the op-supervisor cannot accidentally skip prior sync work, and this way we avoid some previous edge-cases around an empty database. (the panic when the parent block of the first newly inserted block doesn't exist).

Tests

Updated existing tests, and added a test-case to check if the initialization part of the interop deriver works.

[axelKingsley]

Nice, I like this. Now, when we attempt to push data into the database, it does mean that if the supervisor and op-node start correctly, they'll go through one round of error back-and-forth. But as that's just for seeding the supervisor, no big deal.

I added a commit to fix linting, but there's still issues in the Action tests: the action tests assume you can SyncCrossSafe on the chains before doing any work. This triggers the underlying workflow which causes the uninitialized error. So, I commented out the Sync lines, and put them back below a call to InitializeCrossSafe. However, commenting out SyncCrossSafe causes require calls to fail when checking that status.crossUnsafeL2.ID() is correct.

I find this action test failure very surprising, as I wouldn't expect SyncCrossSafe to have any implications for the unsafe. I'll think on this more, but may leave this PR as-is.

@protolambda I'll pre-approve this PR given that the change is sound, and if you manage to fix the action tests then you can go straight to merge.