Separate liveness, readiness and startup services

[richard-axual]

After testing with the ready service for liveness, readiness and startup it became clear that all have different definitions in KSML.

Added the CREATED state to the ComponentState enum.

ProducerRunner now uses states CREATED->STARTING->STARTED->STOPPING->STOPPED or FAILED in case of failures
StreamsRunner now maps the internal Streams state

• KafkaStreams.State CREATED -> ComponentState.CREATED;
• KafkaStreams.State REBALANCING -> ComponentState.STARTING;
• KafkaStreams.State RUNNING -> ComponentState.STARTED;
• KafkaStreams.State PENDING_SHUTDOWN -> ComponentState.STOPPING;
• KafkaStreams.State NOT_RUNNING -> ComponentState.STOPPED;
• KafkaStreams.State PENDING_ERROR, ERROR -> ComponentState.FAILED;

Startup service

• Returns HTTP Status 500 if both ProducerRunner AND StreamsRunner are in state NOT_APPLICABLE (misconfiguration)
• Returns HTTP Status 204 if ProducerRunner AND StreamsRunner is in state NOT_APPLICABLE, STARTING, STARTED, STOPPING, STOPPED
• Returns HTTP Status 500 in other scenarios

Readiness service

• Returns HTTP Status 500 if both ProducerRunner AND StreamsRunner are in state NOT_APPLICABLE (misconfiguration)
• Returns HTTP Status 204 if ProducerRunner is in state NOT_APPLICABLE, STARTED, STOPPING, STOPPED and StreamsRunner is in state NOT_APPLICABLE or STARTED (allows for one to be disabled and the other started, or the ProducerRunner to gracefully stop)
• Returns HTTP Status 500 in other scenarios

Liveness service

• Returns HTTP Status 500 if both ProducerRunner AND StreamsRunner are in state NOT_APPLICABLE (misconfiguration)
• Returns HTTP Status 500 if ProducerRunner is in state FAILED OR StreamsRunner is in state FAILED or STOPPED
• Returns HTTP Status 204 in other scenarios