fix: monitoring UI improvements, poller prompt quality, and timestamp fixes

[AhmadHammad21]

Summary

• Monitoring alert detail now shows Evidence (bullet list), Mitigation Steps (numbered list with colour-coded badges), matching the Slack/Telegram message format
• Monitoring cards show the dedup key resource name (e.g. opendevops-test-failure) so duplicate entries are identifiable at a glance
• Poller prompts upgraded to match event consumer quality — step-by-step instructions, aggregate alarm hint for Lambda namespace, guidance to pull CloudWatch logs and find actual error message
• UTC timestamp fix — SQLite stores timestamps without a Z suffix; frontend now appends it before parsing so timeAgo(), toLocaleString(), and sidebar/history relative times all show correct local time (was 3 hours off for UTC+3)
• test_pipeline.py gains --error-message flag to inject a custom error into the synthetic SQS event, useful for testing different dedup key paths
• evidence column added to alerts table (migration 013); stored and returned by both SQLite and Postgres backends; older rows show empty evidence gracefully

Test plan

• Trigger a test event via scripts/test_pipeline.py — alert detail page shows Evidence, numbered Mitigation Steps, and green/amber/red sections matching Slack message layout
• Monitoring list cards show a small grey monospace resource name below the error text
• Run two test events with different --error-message values — both appear separately on /monitoring
• With SQLite backend: restart server — evidence column added automatically, no DB wipe needed
• Timestamps on monitoring page, sidebar, and history page show correct local time