Interim mitigation for resilience relay bind: timeout bump + fixture retry

[ksuderman]

Refs #469 #470

Interim mitigation for the intermittent Resilience Suite bind-timeout error
(TimeoutError: Pulsar did not bind <mode> consumers) This is a stopgap / safety net; the deterministic root-cause fix is proposed separately in the #470. The two are
non-overlapping and can land in either order.

Changes

• harness/pulsar_control.py: bump the consumer bind timeout from 60.0 to 120.0 to give
  a slow startup more margin.
• conftest.py: _start_fresh_with_retry(). The pulsar fixture retries a fresh
  start once on the bind TimeoutError. A blanket setup retry is more robust than
  marking individual scenarios rerun-eligible, since the error is not pinned to a fixed
  test. Each attempt force-recreates the container (fresh=True), so a retry starts from
  a clean slate. Only the bind TimeoutError is retried — a docker compose up failure
  still raises immediately, and the final TimeoutError propagates if every attempt
  times out so a genuine breakage still surfaces.

No new test dependency (avoids pulling in pytest-rerunfailures); retries silently to
match the suite's existing no-logging style.

Testing

• py_compile and flake8 (max-line 150, complexity 14) pass.
• Not run against the live docker-compose stack.

🤖 Generated with Claude Code