Rework service tasks

[mtitov]

Based on Andre's proposal

After discussion we plan to implement the following schema

• services are defined as normal RP tasks (special task type though)
• service tasks are tagged to a specific pilot instance by setting the pilot field in the TD
• they are submitted via the tmgr
• we introduce a new control message which signals the event that a service task is ready (which implies it has reached AGENT_EXECUTING state. That message type will have the fwd flag set by default
• the tmgr will collect those events and will issue ready-callbacks which are definable on service task types

#3006 (comment)

[mtitov]

Should be considered #2899 - approach is to use tags (exclusive per node)

[andre-merzky]

After this set of PRs is merged, we will consider the following branches for PRs/merges to progress on service tasks and related matters:

• feature/nodb_and_tracer
• feature/sched_reconfig
• feature/service_startup
• feature/zmq_events
• fix/scheduler

These branches all improve the agent service management, profiling (including tapping into the profiling by those services), and adaptive task scheduling (the scheduler also tapping into the profiles and/or getting data from agent services).