add: Error Handling Environment Bundle to Bento configuration

[gregfurman]

Motivation

We need global config to allow for a processing step to return and log an error, with the subsequent processors not being run.

Related to #166 -- where we opted for nacking at the failed processor instead of at the output layer.

Can be used as follows:

pipeline:
  processors:
    - mapping: |
        root = throw(this.value)
    - mapping: |
        root.message = "I'm never reached..."

error_handling:
  strategy: reject

Changes

• Creates config for a new error_handling in the top-level that can be one of reject or none
• Adds an interface that iterates through all messages from a processor and checks if any contain an error.
• Failures will be nacked at the input layer without needing to reach the output.
• Fixes the generate input component always enabling auto_replay_nacks.

TODO

• Add more tests.
• Populate this config at the manager level to allow for nested input.processors and output.processors to also have their results checked for errors.
• Look into adding backoff for retries and shutting off the input to stop new data being read in.
• Add documentation for this feature.

[jem-davies]

I would just change the key strict_mode to something like on_error with the values being an enum of, continue (default) / strict_mode.

So that we can add new modes later on if required, or to add a new mode to change the retry behaviour later? i.e. a new on_error: exponential_backoff.

I think that adding this field to the manager level also makes sense when running in streams mode, it will be ever stream that is created will have the same strict_mode / on_error mode - will simplify understanding?