flaky tls e2e tests: tls_valid_cert_connects_successfully, mtls_no_client_cert_rejected, mtls_valid_client_cert_succeeds

[vieiralucas]

Problem

Three TLS-related e2e tests are flaky in CI — they fail intermittently with connection refused / server not reachable timeouts:

• tls_valid_cert_connects_successfully (crates/fila-e2e/tests/tls.rs)
• mtls_no_client_cert_rejected (crates/fila-e2e/tests/tls_mtls.rs)
• mtls_valid_client_cert_succeeds (crates/fila-e2e/tests/tls_mtls.rs)

Symptoms

All three tests spawn a fila-server process with TLS config and try to connect. Failures are:

• "Connection refused" (server not listening yet)
• "server did not become reachable at 127.0.0.1:XXXXX within 10s"

Likely cause

Race condition between server startup and client connection attempt, possibly exacerbated by CI resource contention. The 10s readiness timeout may be insufficient in slow CI environments, or port binding is intermittently failing.

Observed in

• PR feat: 20.1 binary protocol server — hot-path operations #158 (Epic 20.1) — noted as pre-existing, not caused by PR changes
• Passes on main (intermittent)