Determine the downstream reliance on `TriggerHardForkAtEpoch`

[nfrisby]

As far as we know, this is only used for a single optimization: skipping an initial prefix of HFC eras. (Edit: that assumption was wrong! See the comments below.)

We anticipate an alternative way to achieve that optimization, and so we'd like to remove TriggerHardForkAtEpoch, since it adds an otherwise unnecessary degree of expressiveness to the HFC's semantics---things would be much easier to reason about without it (or at the very least if we sharply narrowed its contract.) But we can simply begin by asking whether anyone is using it for anything other than that optimization.

[nfrisby]

I asked this question in the node channel on Slack:

• in the node channel https://input-output-rnd.slack.com/archives/CR599HMFX/p1696356728786449
• in the CoreTech tribe channel https://input-output-rnd.slack.com/archives/C03T9003RK5/p1696356774042989

[nfrisby]

In a Slack thread https://input-output-rnd.slack.com/archives/CR599HMFX/p1696356728786449, Martin Kourim explained that a test is spinning up a testnet that uses "TestShelleyHardForkAtEpoch": 1 in order to end Byron after one epoch.

I explained there that I think they could use Byron governance instead, and sketched a recommendation for it. But that is some more non-trivial complexity that would be required before we could remove TriggerHardForkAtEpoch.

[nfrisby]

In a Slack thread https://input-output-rnd.slack.com/archives/CR599HMFX/p1696356728786449, Michael Karg confirmed the Benchmarking Team uses this option, but only ever with a value of 0.

[nfrisby]

Closing. It's a week later now. We've heard from the usual suspects and so we're concluding that we have all the info.

Summary: we expected everyone was always using 0. Martin Kourim's team is using 1 in a single use case, but only as a short-cut. It'll take a little elbow grease (from them and/or us) to change that usage to TriggerHardForkAtVersion instead, and it'll be a slightly heavier maintenance burden going forwarding for them, but that seems worthwhile to enable this simplification of the HFC's internal coherency.

[nfrisby]

In the Slack thread https://input-output-rnd.slack.com/archives/CR599HMFX/p1696356728786449), Kostas Dermentzis reported that they have a test using a value of 2, similar to #395 (comment) (although this use is transitioning out of later Shelley eras instead of Byron---so that makes this slightly simpler).