replication_backlog_time not reporting a zero value when node has no writes to send

[cap10morgan]

Steps to reproduce:

1. Setup a Harper cluster w/ this plugin enabled
2. Scrape prometheus metrics from this plugin's endpoint

Expected:

With no more writes to replicate, replication_backlog_time should be reported as 0.

Observed:

replication_backlog_time is > 0.

[cap10morgan]

Additional context from @hdbjeff:

they did a warm up and now all the replication times are good. so def its a bug with something about no pending transactions.

you can see now the time is "stuck " at 500ms and 400ms instead of the hours it was before. but it is still stuck. whereas the DB with transactions flowing shows accurate updates