Fix Atlas Prometheus queries and pantheon PGCR hash resolution

[owens1127]

Summary

• Clamp Atlas metrics interval to at least 1 minute before building PromQL rate windows, preventing invalid [0m] queries that returned HTTP 400 from Prometheus (Fixes SERVICES-8, SERVICES-9).
• Resolve pantheon featured-reprise PGCRs using activityDetails.referenceId when Bungie reports a playlist wrapper in directorActivityHash (Fixes SERVICES-2C).

Root cause (SERVICES-2C)

Pantheon featured-reprise playlists (e.g. hash 153253948) report the playlist wrapper as directorActivityHash but the actual encounter (e.g. Argos 796488315) in referenceId. We were storing the wrapper hash, which is not in activity_version.

Sentry issues addressed

Issue	Events	Fix
SERVICES-8 / SERVICES-9	~450	GetMetrics() clamps intervalMinutes <= 0 to 1
SERVICES-2C	~1000	resolveInstanceActivityHash() prefers referenceId

Deploy

• Merge PR
• ssh raidhub → git pull → make hermes → sudo systemctl restart hermes
• Restart Atlas if running as separate service
• No Postgres migration needed

Test plan

• go build ./apps/atlas/... ./lib/services/pgcr_processing/...
• CI green
• After deploy: SERVICES-2C / SERVICES-8 should stop regressing; reprocess a missed pantheon PGCR if needed

Made with Cursor