fix: bump e2e_pump_until's timeout, tmux test still timed out at 8s
test_e2e_tmux_secondary_da_no_corruption timed out waiting for tmux's
nested shell prompt in two consecutive real ci-tron runs (build/test/
headless all executing at once on a small shared host), even after the
readline-echo corruption fix — a different failure mode (a clean
timeout, not garbled content), consistent with e2e_pump_until's own
doc comment on CPU contention rather than a logic bug. tmux forks a
server *and* a client on top of the one shell every other e2e test
spawns, so it plausibly needs more of the same headroom under real
contention than 8s gave it. Bumped to 20s.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01KpDBFeBQKx9f5U3pfVeVV3