Fold objects by_sqn continues to fold after the Journal snapshot SQN

[martinsumner]

Although the fold will not any objects after the Journal snapshot SQN:

leveled/src/leveled_runner.erl

Lines 363 to 372 in fe9710b

	IsValidFun =
	fun(Bucket, Key, SQN) ->
	LedgerKey = leveled_codec:to_objectkey(Bucket, Key, Tag),
	CheckSQN =
	leveled_penciller:pcl_checksequencenumber(
	LedgerSnapshot, LedgerKey, SQN),
	% Need to check that we have not folded past the point
	% at which the snapshot was taken
	(JournalSQN >= SQN) and (CheckSQN == current)
	end,

The fold will continue to load batches until the end of the last CDB file in the manifest. This may cover > 100K objects, and so the end of the fold is delayed.

This caused a Riak issue with handoffs where the remainder after the last batch is only returned to the riak_core_handoff_sender after the fold has completed - and if this more than 5 minutes after the last batch was sent, the riak_core_handoff_receiver will have timed out, and the handoff does not now complete. The remainder is not received, and if it is an ownership handoff it has to be started from scratch.

[martinsumner]

#466