fix(gateway): bound failed completion notification retries

[KoNit-K]

What does this PR do?

Background-process completion notifications no longer retry forever when the gateway cannot admit their synthetic event. The process watcher now makes at most three admission attempts and doubles the retry delay up to 60 seconds. Successful admissions and the existing completion delivery path are unchanged.

Related Issue

Fixes #138940

Type of Change

• Bug fix (non-breaking change that fixes an issue)

Changes Made

• gateway/run_notifications.py — bounds refused completion admissions and adds capped exponential retry delays for exited background processes.
• tests/gateway/test_background_process_notifications.py — proves that failed admissions stop after the third attempt instead of polling an exited process indefinitely.

How to Test

• scripts/run_tests.sh tests/gateway/test_background_process_notifications.py -q — 31 passed.
• $HERMES_PYTHON -m ruff check gateway/run_notifications.py tests/gateway/test_background_process_notifications.py — passed.

Evidence

• BEFORE RED: the new refused-admission case made 4 delivery attempts; its assertion expected 3.
• AFTER GREEN: scripts/run_tests.sh tests/gateway/test_background_process_notifications.py -q completed with 31 passed.
• CONTROL: Ruff passed for both changed Python files, and successful completion admission remains on the original one-attempt path.

Checklist

Code

• I've read the Contributing Guide.
• My commit message follows Conventional Commits.
• I searched for existing PRs to make sure this is not a duplicate.
• My PR contains only changes related to this fix.
• I've run relevant tests locally (see How to Test).
• I've added a regression test for the retry limit.
• I've tested on my platform: macOS.

Documentation & Housekeeping

• Documentation update: N/A.
• cli-config.yaml.example: N/A.
• CONTRIBUTING.md or AGENTS.md: N/A.
• Cross-platform impact considered; retry timing uses asyncio and is platform-neutral.
• Tool descriptions/schemas: N/A.

[kshitijk4poor]

Thanks @KoNit-K. Your commit is carried into #139026 with your authorship (fcb7baa5c0), so I'm closing this in favour of it.

Why this PR couldn't land on its own: the refusal behind #138940 is always the same admission_conflict (the synthetic wake clears source.message_id, while Discord and Telegram keep it on the committed source). Capping retries stopped the loop, but the completion was still never delivered. A temporary refusal (adapter reconnecting, store draining) also got dropped after about 35 s, where main would have delivered it on the next tick. #139026 adds the root-cause fix from #138967, raises the cap to 8 retries with backoff, and falls back to the plain completion notice instead of break.

#139026 is armed. I'll post the merge SHA here once it lands, and reopen this if it doesn't.

[kshitijk4poor]

Landed on main via #139026 (merge head 2bc1edbc5e). Thanks again.