Adding a retry logic for Hubspot objects loading

[PascalCrow]

This PR is linked to that PR where a retry logic for connection errors was introduced for the loading of associations. This PR now introduces the same logic for the get request of objects to make the loading more stable.

[gemma-claude-assistant]

PR #319 Review — HubSpot retry logic fix

This PR makes two targeted improvements to retry_request in ewah/hooks/hubspot.py:

1. Bug fix: The original code had sleep(60) hardcoded, ignoring the wait_for_seconds parameter entirely. The new code correctly uses sleep(wait_for_seconds).
2. New handling: ConnectionError exceptions are now caught and retried (with a re-raise on the final attempt), matching the intent of the existing retry loop.
3. Default reduced: wait_for_seconds default halved from 60 → 30 seconds, consistent with the corrected sleep usage.
4. Minor improvement: Sleep is now skipped on the last failed attempt (if try_number < retries), avoiding a pointless wait before giving up.

The logic is correct: a ConnectionError on the final attempt re-raises before reaching return request, so there's no risk of NameError on an unbound variable. Version bump to 0.9.27 is present as required.

Approved — no blocking issues.

[gemma-claude-assistant]

Claude review cost: $0.1846 · 4 turns · 90s · eu.anthropic.claude-sonnet-4-6