e2e: add settling checks between liveness cases to fix intermittent flake

[juan-malbeclabs]

Summary of Changes

• After unblocking a client in the route liveness matrix, the test now waits for all expected-stable routes to converge before starting the next block/unblock cycle
• Specifically: after Case B unblocks client2, the test explicitly verifies c1->c3 and c3->c1 are stable before Case C begins blocking client3
• Without this, BGP propagation from the previous cycle could still be in-flight when Case C starts, causing the c1->c3 restored assertion to race against the 60s timeout

Diff Breakdown

Category	Files	Lines (+/-)	Net
Tests	1	+6 / -2	+4

Single-file change — pure test stabilization, no production code touched.

Key files (click to expand)

• e2e/multi_client_ibrl_liveness_test.go — adds requireEventuallyRoute settling checks for c1->c3 and c3->c1 after Case B's unblock, and adds a comment explaining why c1->c4 is intentionally omitted (client4 has liveness disabled and never responds to probes)

Testing Verification

• Ran TestE2E_MultiClientIBRL_RouteLiveness with -count=3 to exercise three consecutive passes; all passed cleanly (~366–407s per run)
• Confirmed that adding a c1->c4 restored check (the naive approach) fails consistently, validating the understanding that client4's liveness-disabled config prevents route restoration via probe responses