op-batcher: Fix max channel duration tracking for large block backlogs

[sebastianst]

The batcher supports setting a max channel duration that controls for long (in terms of L1 blocks) to keep open a channel. This effectively enables chain operators to specify a max duration between safe head increases.

However, the way this is currently implemented it also applies when the batcher is working its way through a large backlog of blocks. During a recent safe head lagging incident on OPM on Oct 10/11, during which the batcher was throttling for a prolonged period of ~2h, this caused the batcher to create only half-full 1-blob channels that span the configured duration of ~15 min.

We need to fix this so that the max duration is not applied when working through a block backlog.