Expose more accurate task / run lifecycle via Client API

[tuulos]

Today, you can use the Client API to see whether a task/run has completed successfully (including caught exceptions) through .finished and .successful attributes.

Through the current API, you are not able to see the state of runs that

• are scheduled or being started by a production orchestrator like Step Functions or Argo Workflows
• have crashed vs. that are still running.

It would be useful to expose more accurate task/run state information through the Client API, some of which is already visible in the Metaflow UI (through heartbeats).

[jixianyihao]

This feature we really need， like spark connect server， https://spark.apache.org/docs/latest/spark-connect-overview.html spark provide API make IDE or Jupyterlab can submit job use API； It is very important。