terraform-provider-azapi feature request: user configurable retryable errors

Hi,

I would like to propose that the resources contained within this provider have an additional schema property to identify retryable errors.

With any complex cloud infrastructure, retryable errors are a fact of life. We could make our customers' lives easier by allowing them (and module authors) the ability to configure retries.

Inspiration: https://pkg.go.dev/github.com/cenkalti/backoff

Proposed configuration

resource "azapi_resource" "example" {
  # usual resource properties

  # Retry configuration
  retry {
    interval_seconds     = 5    # mandatory
    randomization_factor = 0.5  # optional, default is 0.5
    multiplier           = 2    # optional, default is 2
    error_message_regex  = [
      "my error message",
      "end of error message$"
    ]
  }

  # Retry will continue up to configured timeouts
  timeouts {
    create = "10m"
  }
}

Nov 10 '23 12:11 matt-FFFFFF

@lonegunmanb would be great to get your thoughts on this based on our recent conversations!

Nov 10 '23 12:11 matt-FFFFFF

Thanks for asking @matt-FFFFFF , I would like to add max_attempts and max_interval.

max_attempts is easy to understand, but if the multiplier is big, it might lead a too long interval that makes this exponential backoff unreasonable, in that case would a max_interval solve this issue?