Incorrect "total" value in search in nested index

We have deduplication for list of IDS

if total == 0 || lastID != id { // lids increase monotonically, it's enough to compare current id with the last one
	ids = append(ids, seq.IDSource{ID: id})
}
lastID = id

But we didn't have it for total calculation. To use deduplication logic, we need to convert LID to seq.ID, this requires reading MIDs and RIDs blocks from disk, which slows down the query.

An additional problem is that in the context of performing a search, we have no information that a nested index is involved in the search, and we cannot decide to take the more expensive branch for the "total".

See iterateEvalTree() function in frac/processor/search.go

Start with iterateEvalTree() in frac/processor/search.go and trace how total is calculated versus the existing deduplication of IDs. Done means nested-index searches report a deduplicated total while accounting for the additional disk reads required to convert LIDs to seq.IDs.