Add estimate of remaining time for indexer

[nrejac]

It would be useful to have an estimate of time remaining when running the indexer. Currently only the rate of indexing (records per second) is reported, but if possible, the estimated remaining time based on rate * total number of records to index would be helpful.

[danstoner]

We don't have a "count" anywhere.

For performance we are fetching from cursors and iterators which don't know the size until they get to the end (I think).

We could probably add an SQL count before we begin work but in postgres count is notoriously slow.

[danstoner]

postgres keeps table statistics (for query planner) that have numbers that are generally going to be on the same order of magnitude as the table count.... so that might work for a time estimate.

Actually, with newer postgres features such as Parallel Seq Scan, the time to compute count(*) is much faster than before, definitely fast enough to work for this use case.