Add configuration parameter for output retries attempts

## Feature Request

Currently, the output retry attempt is done only once after a 15-second sleep and then exits if the output fails. Users would need the option to configure the number of 'retries' for when outputs fail.

### Proposal:
Introduce a new `[agent]` parameter that can configure the number of retries for outputs.

### Current behaviour:
Telegraf retries once after a 15s sleep and then exits with an error if it is not able the get an 'output' plugin to work.

### Desired behaviour:
Telegraf retries for the configured number of attempts by the user before it exits.

### Use case:
Scenarios where monitoring stack or outputs, in general, take time to startup or are deployed after the dependant component in question.
Telegraf must be configurable for retries

Start by tracing Telegraf's agent output retry handling and the existing [agent] configuration parameters. Determine where the current single retry and 15-second sleep are implemented, then review related tests if present. Done means users can configure the retry attempt count and Telegraf retries that many times before exiting on output failure.

Written by the indexing model from the issue text.