Request tracing support across core and connectors

[lahirujayathilake]

Summary

Adds a single trace ID to every audit row a request produces. The trace ID flows through the core service, the event bus, and the AMIE, COmanage, and SLURM connectors. Also adds a small set of admin endpoints under /audit/* so operators can look up a trace, see the full event chain, and filter by subsystem, status, or time window.

Before this change, when a multi-step request failed somewhere (for example, an AMIE provisioning request that triggered downstream COmanage work), audit rows existed but had no link between them. There was no way to figure out "what was the full chain of events for this request, and where did it break?". This PR fixes that.

How it works

Audit rows are written by the components that do the work. AMIE writes structured rows as it walks each incoming packet through its state machine. The COmanage and SLURM connectors write rows when their subscribers start handling an event, and again on every success or failure outcome. Every row carries a trace ID, a span ID, a parent span ID, and a source label (core, amie, comanage, or slurm). The IDs let the admin endpoints rebuild the event tree. The source label is used for filtering.

Every audit row in Custos lives in one audit_events table. Core, COmanage, SLURM, and AMIE all write to it with the same shape, tagged by source. AMIE used to keep its own amie_audit_log table with foreign keys into AMIE-specific tables. That is gone in this PR. The AMIE-specific references it still needs (the originating packet and the per-attempt processing event) now live in a small table called amie_audit_extras, joined to audit_events on audit_event_id. The trace viewing layer reads audit_events directly, no UNION, and the source label tells operators which subsystem produced each row.

The shape generalizes. Any future connector that needs to attach its own references to an audit row creates a <connector>_audit_extras table and a matching /connectors/{name}/... endpoint that joins it. For now there is one connector-specific endpoint, GET /connectors/amie/packets/{packet_id}/audits, which returns every audit row written for one AMIE packet. The core audit_events table stays neutral and never grows connector-shaped columns.

The admin endpoints are read-only. They let you list traces with filters, fetch the event tree for one trace, fetch a flat time-ordered list for a trace or a single span, and list the available subsystems for the filter UI.

[DImuthuUpe]

We need to closely look at the data models again the verify whether we can compress those further

[DImuthuUpe]

@lahirujayathilake please fix merge conflicts

[lahirujayathilake]

@lahirujayathilake please fix merge conflicts

@DImuthuUpe, I updated the PR