[packaging][centos] Improve service (systemd + init script)

[olivielpeau]

What does this PR do?

• Makes agent service depend on remote_fs and network services on CentOS (LSB headers): this should hopefully give more guarantees that the agent service doesn't start up too early in the boot process.
• Move systemd service files up one directory, required by [centos] Improve systemd and service restart behaviors dd-agent-omnibus#194 : that PR makes these files common to debian/centos/suse packages.

Testing Guidelines

These changes pass dd-agent-testing.

[truthbk]

[packaging][centos] Improve service (systemd + init script)

[hegyre]

Hello,

Is there a possibility to have the reload command working ?

Currently with v5.17, we can only use the restart command..
Old behavior (with version <= 5.16) acted like this when we issued a reload:

systemctl reload datadog-agent.service

systemctl status datadog-agent.service
Sep 01 00:30:10 dcsjp-web-qas-to-i-076fad0601d1bd713.jp.pmi-dcs.com dd.collector[2322]: INFO (agent.py:123): SIGHUP caught! Scheduling configuration reload... run.
Sep 01 00:30:10 dcsjp-web-qas-to-i-076fad0601d1bd713.jp.pmi-dcs.com systemd[1]: Reloaded LSB: Start and start datadog-agent.

With version 5.17, the behavior is now:

systemctl reload datadog-agent.service

Failed to reload datadog-agent.service: Job type reload is not applicable for unit datadog-agent.service.
See system logs and 'systemctl status datadog-agent.service' for details.

Thanks in advance.