Improving bot processing metric flow

[aomerk]

• Aligning all processXXX methods Due to a difference in error handling, block/tx processors aren't submitting *.request and *.latency metrics when there are invocation errors

• Starts submitting error metrics for invocation errors This is different from *.error metrics we have right now. Instead of having error metrics only when there is a client-side error (ie. exceptions, error responses), nodes will emit metrics for invocation errors as well

Following is the request processing lifecycle metric flow:

- *.request
    - every attempt

- *.latency
    - every request has a latency

- *.success
    - response type success

- *.error
    - response type error
    - err != nil (invoke error)
        - details: err.Error()