Panic when shutting down (and a batch is being built)

[hsanjuan]

2022-12-03T17:14:44.428Z	INFO	cluster	ipfs-cluster/cluster.go:794	shutting down Cluster
2022-12-03T17:14:44.428Z	INFO	restapi	common/api.go:501	stopping Cluster API
2022-12-03T17:14:44.428Z	INFO	pinsvcapi	common/api.go:501	stopping Cluster API
2022-12-03T17:14:44.428Z	INFO	ipfsproxy	ipfsproxy/ipfsproxy.go:297	stopping IPFS Proxy
2022-12-03T17:14:44.428Z	INFO	crdt	crdt/consensus.go:349	stopping Consensus component
2022-12-03T17:14:50.697Z	INFO	crdt	go-ds-crdt@v0.3.9/set.go:122	Tombstones have bloomed: 23711484 tombs. Took: 26.266125565s
2022-12-03T17:14:50.730Z	INFO	crdt	go-ds-crdt@v0.3.9/crdt.go:296	crdt Datastore created. Number of heads: 1. Current max-height: 845951. Dirty: true
2022-12-03T17:14:50.731Z	INFO	crdt	crdt/consensus.go:319	'trust all' mode enabled. Any peer in the cluster can modify the pinset.
2022-12-03T17:14:50.731Z	INFO	crdt	crdt/consensus.go:324	'crdt batching' enabled: 1000 items / 1m0s
2022-12-03T17:14:50.731Z	WARN	crdt	go-ds-crdt@v0.3.9/crdt.go:507	store is marked dirty. Starting DAG repair operation
panic: runtime error: invalid memory address or nil pointer dereference
[signal SIGSEGV: segmentation violation code=0x1 addr=0x30 pc=0x7217f0]

goroutine 3294 [running]:
go.opencensus.io/trace.(*tracer).FromContext(...)
	go.opencensus.io@v0.23.0/trace/trace.go:117
go.opencensus.io/trace.(*tracer).StartSpan(0x5a59444950594d4c?, {0x0, 0x0}, {0x16bb0a5, 0x10}, {0x0, 0x0, 0x5a505a4d574b5935?})
	go.opencensus.io@v0.23.0/trace/trace.go:176 +0x90
go.opencensus.io/trace.StartSpan(...)
	go.opencensus.io@v0.23.0/trace/trace_api.go:56
github.com/ipfs-cluster/ipfs-cluster/state/dsstate.(*State).Rm(0xc079473b80, {0x0, 0x0}, {{{0x0?, 0x44f365?}}})
	github.com/ipfs-cluster/ipfs-cluster/state/dsstate/datastore.go:100 +0x97
github.com/ipfs-cluster/ipfs-cluster/consensus/crdt.(*Consensus).batchWorker.func1({{0x0, 0x0}, 0x0, {{0x0, 0x0, {0x0, 0x0}, 0x0, 0x0, {0x0, ...}, ...}, ...}, ...})
	github.com/ipfs-cluster/ipfs-cluster/consensus/crdt/consensus.go:517 +0xcc
github.com/ipfs-cluster/ipfs-cluster/consensus/crdt.(*Consensus).batchWorker(0xc0004a6500)
	github.com/ipfs-cluster/ipfs-cluster/consensus/crdt/consensus.go:548 +0x29d
created by github.com/ipfs-cluster/ipfs-cluster/consensus/crdt.(*Consensus).setup
	github.com/ipfs-cluster/ipfs-cluster/consensus/crdt/consensus.go:329 +0x8d8

This would be hard to reproduce after c10c37f, but it still should not have happened.