Jaeger: Single lookback_period_hours causes performance vs functionality trade-off

[nico34638]

Describe the bug
The current Jaeger configuration in Quickwit uses a single lookback_period_hours parameter for all time-unbounded queries (get_services, get_operations, and get_trace operations). This creates a problem when you need different time windows for service/operation discovery versus trace retrieval operations.
The issue is that I cannot increase the current parameter because it would make the search page display too slow (service and operation discovery would take too long), but I need the find_trace functionality to work with a longer lookback period because it's very convenient for sharing traces between colleagues via trace links.

Steps to reproduce (if applicable)
Steps to reproduce the behavior:

• Configure Quickwit with Jaeger endpoint enabled and a short lookback_period_hours (e.g., 24 hours) for acceptable UI performance
• Use Jaeger UI to browse services and operations (this works well with short lookback period)
• Try to share a trace link with a colleague for a trace that is older than the configured lookback period
• The trace retrieval will fail or return incomplete results because it's using the same short lookback period
• If you increase lookback_period_hours to fix trace sharing, the search page becomes unacceptably slow

Expected behavior
There should be separate configuration parameters:

• One for service/operation discovery operations (get_services, get_operations) that can be set to a shorter period (e.g., 24-72 hours) for good UI performance
• Another for trace retrieval operations (get_trace, find_traces) that can be set to a longer period (e.g., 30 days) to ensure trace links work properly for sharing traces between colleagues and accessing historical traces

Configuration:
jaeger: enable_endpoint: true lookback_period_hours: 24 # Can't increase this without making search page too slow max_trace_duration_secs: 3600 max_fetch_spans: 10000

Use case impact

• Performance constraint: Cannot increase lookback_period_hours because service/operation discovery becomes too slow
• Collaboration need: Need longer lookback for find_trace to share trace links between colleagues
• Historical analysis: Need to access older traces for debugging and investigation purposes

Proposed solution
Split the lookback_period_hours into two separate parameters:

• service_discovery_lookback_period_hours: For get_services and get_operations (default: 24-72 hours for performance)
• trace_lookback_period_hours: For get_trace and trace retrieval operations (default: longer period, e.g., 30 days for trace sharing)

Maintain backward compatibility by keeping the existing lookback_period_hours as a fallback when the new parameters are not specified.