Logging error should occur once

See: https://github.com/influxdata/telegraf/issues/5964 https://github.com/influxdata/telegraf/pull/6166 https://github.com/influxdata/telegraf/pull/9888

### Proposal:
Error messaging should log once - implement across Telegraf globally. Will require audit of plugins doing it "the wrong way".

### Current behavior:
Every single iteration, if the interval is set to `60s`, Telegraf errors for every interval for services that don't exist.

### Desired behavior:
If the service doesn't exist, I'd expect it to flag the issue on the first stdout as not existing and then refrain from logging as error each time.

### Use case:
Ex: You type the name of the service wrong, and instead of a warning the plugin just doesn't collect anything.

Start by reading issue 5964 and pull requests 6166 and 9888, then inspect how Telegraf plugins currently report recurring errors. The scope is a global audit across plugins; done means the proposed once-only error behavior is consistently defined and applied, with coverage for services that do not exist.

Written by the indexing model from the issue text.