op-batcher: DA metrics not reset when state is cleared

[geoknee]

Under certain conditions, the batcher will clear it's state and "start fresh". However, it is not always resetting metrics which need to stay in sync with the state which is now cleared.

optimism/op-batcher/batcher/channel_manager.go

Lines 76 to 88 in 4d832fa

	// Clear clears the entire state of the channel manager.
	// It is intended to be used before launching op-batcher and after an L2 reorg.
	func (s *channelManager) Clear(l1OriginLastSubmittedChannel eth.BlockID) {
	s.log.Trace("clearing channel manager state")
	s.blocks.Clear()
	s.blockCursor = 0
	s.l1OriginLastSubmittedChannel = l1OriginLastSubmittedChannel
	s.tip = common.Hash{}
	s.currentChannel = nil
	s.channelQueue = nil
	s.metr.RecordChannelQueueLength(0)
	s.txChannels = make(map[string]*channel)
	}

The channel queue length metric is correctly reset, but we need to do something similar for pendingDABytes, otherwise the batcher may needlessly be throttling the sequencer.