JIT EBBs in the ThreadNet tests require ExtLedgerState in block production

[nfrisby]

Fixes #2047.

PR #1942 changed blockProduction to take a LedgerState instead of
ExtLedgerState, which suffices for all forging. However, the ThreadNet tests'
current approach to EBBs requires anticipating during block production whether
the produced block will be invalid in anyway. (The node does not emit the EBB
unless the EBB unless it is also emitting a valid block.) And that anticipatory
validation must include protocol-level errors, which requires the full
ExtLedgerState.

It may be possible to instead change the tests to allow for EBBs without
successors, but I'm don't see how, so I currently propose leaving that for future work if we
prioritize it.

Edit: PR 2176 essentially fixed this failure mode en passant. This PR now just simplifies the JIT EBB logic accordingly.

[nfrisby]

@edsko and I reviewed on a call. Thanks!

Plan: re-expanding the produceBlock API as this PR currently does is undesirable. Edsko saw alignment with other concerns and so pressed for the new CannotLead mechanism (see merged PR #2169). I'm going to rebase this PR to rely on that instead; the goal is that the tests should never generate an invalid block because the CannotLead check will abort the thread upstream. Until Issue #2173, this will also involve dialing back the Gen NodeJoinPlan so that it avoids all PBftExceededSignThreshold errors.

[nfrisby]

Once PR #2176 is merged, this PR will change dramatically; it will merely simplify the JIT EBB generator, under the assumption that (at least in the test suite for now) nodes now never forge invalid blocks!

[edsko]

Just marking this as changes required. The changes are already described by @nfrisby himself, just want to avoid an accidental merge.

[edsko]

Nice, glad that that worked out!

[edsko]

bors merge

[iohk-bors]

Build succeeded

• ci/hydra:Cardano:ouroboros-network:required
• buildkite/ouroboros-network