Harden LongBridge runtime errors and notification handling

[Pigbibi]

Why

• Runtime market-hours checks could report a warning even when the provider returned (open, None).
• Long error tracebacks were sent through compact notification channels, causing noisy or rejected messages.
• Execution-report persistence needed to tolerate both the current cloud URI result and the legacy GCS URI field.

What changed

• Emit market-hours warnings only when an actual error is present.
• Keep detailed tracebacks in durable reports while sending bounded compact notifications.
• Normalize execution-report persistence return fields across dependency versions.
• Keep Cloud Run validation routes non-mutating and update compatibility fixtures/regression coverage.

Verification

• pytest: 287 passed, 1 warning; 15 subtests passed.
• ruff check: passed for all changed Python files.
• git diff --check: passed.
• Codex review (not AIAudit): no actionable findings.

Production safety

• The credential-refresh persistence permission was repaired separately and the current valid credential was retained.
• Post-repair service checks were healthy; no live order endpoint was invoked.