feat: added capability to extract metrics from the logs

[CptSchnitz]

Main feature: derive Prometheus metrics from nginx access logs via Fluent Bit, instead of just shipping raw logs.

• OGC request parsing (Lua) — new custom.lua script hooked into Fluent Bit that parses both KVP (?service=WMS&request=...) and RESTful (/wmts/.../{z}/{x}/{y}.{fmt}) URLs to extract ogc_service, ogc_operation, ogc_layer, ogc_zoom (bucketed into ogc_zoom_group to control cardinality), and ogc_format, plus a normalized result (ok/error) and request_time_seconds.
• Metrics generation (log_to_metrics filters) — new Fluent Bit filters that turn the parsed log fields into:
  • mapproxy_wms_requests_total / mapproxy_wmts_requests_total counters (by operation, layer, format, result — WMTS also labeled by client name)
  • nginx_http_errors_total counter for all 4xx/5xx responses
  • mapproxy_{wms,wmts}_request_duration_seconds latency histograms (labeled by result)
• Simplified log format — dropped the large, verbose log_format.conf (nginx log_format directive with ~50 fields) since raw stdout logs are no longer needed now that metrics are extracted (chore: remove stdout).
• Dependency bump — nginx subchart upgraded 2.3.1 → 2.4.0 to pick up the fluentbit/log_to_metrics/Lua filter support this feature relies on.

Net effect: mapproxy nginx now exports structured request/latency metrics (by OGC service, operation, layer, zoom bucket, format, and error status) directly from access logs, rather than relying on log aggregation alone.