telemetry/bgpstatus: collect BGP state from all tenant VRF namespaces

[juan-malbeclabs]

Summary of Changes

• Users whose tenant has VrfId != 1 have their GRE tunnel interface placed in ns-vrf<N> on the Arista device, but the BGP status submitter was only checking ns-vrf1. This caused a persistent "tunnel not found" debug log and left those users' onchain BGP status permanently stale.
• Fix: on each tick, derive the full set of VRF namespaces from programData.Tenants via a new vrfNamespaces helper, then collect BGP socket stats and local interfaces from all of them before the per-user loop. Tunnel IPs are globally unique (onchain-allocated), so merging across namespaces is safe.
• Introduced NamespaceCollector as an injectable function type in Config, replacing LocalNet. DefaultCollector wraps the real Linux calls; tests supply a mock without any Linux syscalls.

Diff Breakdown

Category	Files	Lines (+/-)	Net
Core logic	2	+82 / -21	+61
Scaffolding	1	+1 / -1	0
Tests	3	+508 / -11	+497

Mostly test additions; the fix itself is compact -- ~60 net lines across two core files.

Key files (click to expand)

• controlplane/telemetry/internal/bgpstatus/submitter.go -- adds DefaultCollector (wraps Linux calls), rewrites tick() to loop over all VRF namespaces and merge results; aborts only if every namespace fails
• controlplane/telemetry/internal/bgpstatus/bgpstatus.go -- adds NamespaceCollector func type, vrfNamespaces helper (derives namespace list from tenant VRF IDs), replaces LocalNet with Collector in Config
• controlplane/telemetry/internal/bgpstatus/submitter_linux_test.go -- four new tick() behavioral tests: user in ns-vrf2 found and reported Up; partial namespace failure continues; all-namespace failure aborts
• controlplane/telemetry/internal/bgpstatus/submitter_test.go -- migrates test helpers to NamespaceCollector, adds five vrfNamespaces unit tests (dedup, zero VrfId skip, multi-tenant)
• e2e/user_bgp_status_test.go -- adds TestE2E_UserBGPStatus_NonDefaultTenant: creates a tenant (VrfId != 1), connects a client under it, and verifies Up -> Down BGP status transitions onchain
• controlplane/telemetry/cmd/telemetry/main.go -- wires DefaultCollector(localNet) into bgpstatus.Config

Testing Verification

• All existing bgpstatus unit tests pass unchanged
• TestVrfNamespaces_* (5 cases): deduplication, zero VrfId skip, base namespace always included, additional VRF appended
• TestTick_* (4 cases, Linux): user in ns-vrf2 is found and reported Up; one failing namespace is warned and skipped while others succeed; all namespaces failing aborts the tick with no submissions
• TestE2E_UserBGPStatus_NonDefaultTenant: end-to-end validation that a user in a non-default tenant VRF reaches BGP status Up onchain when the session is established, and Down after the daemon is killed