op-batcher: Watch timeouts [ENG-2625]

Update bridges.md

And use it in op-batcher.

specs: Fix typo & link in withdrawals section

Remove node built in from sdk

feat(mon): Add withdrawal monitor to chain-mon

chore(op-node): Fix two typos

chore(deps): bump http-cache-semantics from 4.1.0 to 4.1.1

fix(spec): Add missing call to engine_newPayload

The batcher now queries the inbox address from the rollup node.

- also fix a captured loop variable

- Removes the ChannelSubTimeout config flag and instead uses the SubSafetyMargin
  as a #L1-block safety margin to both, the channel timeout and sequencing
  window timeout
- Remove unused flags SequencerBatchInboxAddress and ChannelTimeout as
  these values are now queried from the rollup node.

This avoids having to provide rollup configuration as flags to the
batcher; also avoids possible inconsistencies.

We need to access the batch data directly in the batcher when tracking
the sequencer window. For that reason, export BlockToBatch to be able to
create the batch from a block there and then add it to ChannelOut without
converting twice.

and improve other code docs. It was confusing that AddL2Block adds a
block to the queue whereas addBlocks processed them, thereby removing
them from the queue. So renamed to `processBlocks`.

The batcher needs to observe two different channel timeouts.
1. the consensus channel timeout, so it knows when to resubmit a channel
   whose frames didn't land on L1 quickly enough and
2. the channel submission timeout, by which time it closes the channel
   even if the target amount of data hasn't been put into the channel
   yet.

The process to create a new block via the engine API only returns the block and doesn't import it.
So an explicit call to engine_newPayload is required to import the payload before making it canonical.
The actual code performs this call, but it was missed in the spec description.

fix(spec): Withdrawal calls are not forwarded during the proof verification call

The spec incorrectly stated that the call was forwarded at both the proof verification call and during the finalization call.