Update grafana PID storage logic

Fixes #116 (closed)

Previously, the Grafana PID file ($GRAFANA_DATA_DIR/grafana.pid) could become stale or not get updated correctly when:

• Grafana failed to start properly.
• User forgot to run stop-all-swh-backends.sh so the PID file existed before Grafana started
• Grafana crashed or was killed externally, leaving an outdated PID file.

This caused shutdown scripts to fail, and in some cases, Adastra admins had to manually kill orphaned processes.

Changes

• Added a retry mechanism to validate that the Grafana process is running before writing the PID file instead of waiting for the file to exist.
• Ensured the PID is only written after a successful process check (ps + grep grafana).
• Improved logging so the PID lifecycle is clearer.

How to Test

1. Run the whole pipeline on Adastra
2. Confirm that grafana.pid is created only after Grafana is fully started.