triedb/pathdb: fix an issue of history indexer deadloop

[jsvisa]

In our environment, in some circumstances, we saw the below logs:

DEBUG[08-03|03:18:31.894] State history is fully indexed           last=1,075,884
DEBUG[08-03|03:18:31.894] Resume state history indexing            id=1,075,967        tail=929,308
DEBUG[08-03|03:18:31.894] State history is fully indexed           last=1,075,884
DEBUG[08-03|03:18:31.894] Resume state history indexing            id=1,075,967        tail=929,308
DEBUG[08-03|03:18:31.894] State history is fully indexed           last=1,075,884
DEBUG[08-03|03:18:31.894] Resume state history indexing            id=1,075,967        tail=929,308
DEBUG[08-03|03:18:31.894] State history is fully indexed           last=1,075,884
DEBUG[08-03|03:18:31.894] Resume state history indexing            id=1,075,967        tail=929,308
DEBUG[08-03|03:18:31.894] State history is fully indexed           last=1,075,884
DEBUG[08-03|03:18:31.894] Resume state history indexing            id=1,075,967        tail=929,308
DEBUG[08-03|03:18:31.894] State history is fully indexed           last=1,075,884
DEBUG[08-03|03:18:31.895] Resume state history indexing            id=1,075,967        tail=929,308
DEBUG[08-03|03:18:31.895] State history is fully indexed           last=1,075,884
DEBUG[08-03|03:18:31.895] Resume state history indexing            id=1,075,967        tail=929,308
DEBUG[08-03|03:18:31.895] State history is fully indexed           last=1,075,884
DEBUG[08-03|03:18:31.895] Resume state history indexing            id=1,075,967        tail=929,308
DEBUG[08-03|03:18:31.895] State history is fully indexed           last=1,075,884
DEBUG[08-03|03:18:31.895] Resume state history indexing            id=1,075,967        tail=929,308
DEBUG[08-03|03:18:31.895] State history is fully indexed           last=1,075,884
DEBUG[08-03|03:18:31.895] Resume state history indexing            id=1,075,967        tail=929,308

After some digging of the code, I found the below steps:

1. when the beginID > lastID, the indexer will be returned immediately
2. The initer.run checks whether the initialization of the indexer is complete. If it is not, a new indexer is created and run again.
3. However, since lastID is not updated, the condition of beginID > lastID remains true, causing the process to loop back to step 1.

Here we restore the metadata with the lastId(which maybe it's the bottom layer's id). So the initer will be exited later, skip the deadloop.

[rjl493456442]

DEBUG[08-03|03:18:31.895] State history is fully indexed           last=1,075,884
DEBUG[08-03|03:18:31.895] Resume state history indexing            id=1,075,967        tail=929,308

I think the root cause is: the last was not reverted correctly in the chain rollback previously? It's not expected to be higher than state history.