fix: retry gemini_community.py's GitHub calls once on transient network failure

[essentialbit]

Summary

• gemini_community.py duplicates community.py's _gh_get/_gh_post/_get_discussions/_post_discussion_comment helpers but was missed by the identical single-retry-on-transient-failure fixes already shipped for community.py (fix: retry _gh_get once on transient network failure before giving up #584 _gh_get, fix: retry _gh_post/_gh_delete once on transient network failure #588 _gh_post/_gh_delete, fix: retry Discussions GraphQL calls once on transient network failure #592 Discussions GraphQL).
• Applies the same single-retry/2s-delay pattern to all four sibling call sites in this file, closing the gap so Gemini's own community-engagement runner degrades the same way on a transient 5xx/network error instead of silently dropping the read/write.
• gemini_community.py has no _gh_delete equivalent, so nothing else to fix here.

Test plan

• python3 -c "import gemini_community" — imports cleanly
• python3 -c "from main import *" — full app import unaffected
• Scratch mocked-requests script (deleted after use): transient-then-recover for all 4 functions, persistent-failure bounds at 2 attempts, legitimate 200-empty response never spuriously retries — all pass