fix(client): Respect channel timeout

[clabby]

Overview

Currently, we don't respect the channel timeout in kona-client. This change will be a painful one to make for performance, but we need to respect consensus rules rather than batcher policy.

In real-world application, the OP Stack batcher almost always opens and closes channels in a single transaction. However, this is strictly not guaranteed by the protocol.

To accommodate for this, we must start derivation at least RollupConfig::channel_timeout blocks before the L1 origin of the trusted safe head.

https://github.com/ethereum-optimism/kona/blob/e069eb4178af1a2a7ea1f772f3f35c240d178534/bin/client/src/l1/driver.rs#L203-L203

[refcell]

Discussed offline: attempt to derive the channel in the happy path from the l1 origin. If it fails to derive the channel, walk back the channel timeout window of L1 blocks.