Fix flaky PUBSUB pattern tests by making channel names deterministic

[jh1231223]

Fixes #4944

This PR stabilizes the following tests:

• test_pubsub_exact_happy_path_many_channels
• test_sharded_pubsub_many_channels
• test_pubsub_pattern
• test_pubsub_pattern_many_channels

This work follows the same approach used in:

• tests(pubsub): deflake test_pubsub_combined_exact_and_pattern_one_client by making channel names unique per index #4880

---

What I changed

I updated the channel-name construction in the affected tests to make each channel deterministically unique while preserving the original semantics (same shard / same pattern).

Specifically, by embedding the loop index into each channel name:

• It guarantee the intended number of distinct channels per test, regardless of random collisions.
• The tests now always publish and expect the same number of messages, so message-count assertions are stable across runs.

---

Validation

Tested locally with repeated runs:

python -m pytest -c /dev/null -q \
  tests/async_tests/test_pubsub.py \
  -k "[test_name]" \
  --count=600 -s --maxfail=1