Hydra node cannot handle time conversion once 2k blocks have passed

[ghost]

Context & versions

This is a follow-up issue that surfaced as we worked on cleaning up contestation deadline in #483

Steps to reproduce

Note these are all very abstract, we know there is a bug but I did not actually observed it

1. spin up a hydra-node
2. open a head
3. wait for 2k blocks to pass on-chain
4. try to compute contestation period by closing the head

Actual behavior

Node fails to compute contestation deadline (and crashes?)

Expected behavior

Node should correctly compute contestation deadline.

[ghost]

This is a follow-up to #483.

Solutions envisioned:

• Short-term: Do not allow heads to live longer than 2k
• Assume no upper-bound on current Era's time parameters
• Requery EraHistory upon every time computation

[ch1bo]

This is because the time handle queried here becomes to old and we should have addressed this comment: #483 (comment).

[ch1bo]

We discussed the two approaches and decided to:

• Do the morally right thing first by querying the time handle on each block rolled forward using a MonadTimeHandle type class to abstract over IO/IOSim.
• If this is too expensive or not possible, we draw the line there and fall back to use unsafeExtendSafeZone, possibly updating the TimeHandle to fetch once and deliver current time/slot continuously instead.