feat: worktime scoring strategy for Prometheus discovery

[Copilot]

Adds a strategy-pattern–based scoring function that weights Prometheus range query data points by time of day before aggregation. This lets operators prioritize images pulled during business hours over nighttime noise.

Changes

• API types (api/v1alpha1): ScoringStrategy, WorktimeStrategy, WorktimeWindow — configurable time-of-day windows with weight multipliers and timezone support
• Strategy interface (internal/discovery/scoring.go): ScoreWeighter interface with worktimeWeighter implementation; extensible for future strategies
• Aggregation integration: aggregateRangeValues applies per-datapoint weight multiplication when a weighter is configured; no-op when nil (backward compatible)
• Controller wiring: buildSource constructs the weighter from API config and passes it through to PrometheusSource
• CRD + deepcopy regenerated via make codegen

Example

prometheus:
  endpoint: https://mimir.example.com
  lookback: 168h
  aggregationMethod: sum
  scoringStrategy:
    type: worktime
    worktime:
      timezone: "Europe/Berlin"
      windows:
        - { startHour: 9,  endHour: 17, weight: "1.0" }
        - { startHour: 6,  endHour: 9,  weight: "0.3" }
        - { startHour: 17, endHour: 19, weight: "0.3" }

Data points at 3am get weight 0 (dropped from sum); data points at 10am get full weight. The strategy is optional and nil-safe — existing configs are unaffected.