[VRP] Harden testcase safety check in long-lived bots

[PauloVLB]

Problem

Long-lived bots might attempt to handle untrusted testcases locally (e.g., during minimization or analysis) if not properly isolated. The previous safety check relied on the UNTRUSTED_UTASK environment variable, which was unreliably set and often led to false positives or missed detections.

Proposed Solution

Harden the testcase safety check by relying on the rock-solid discriminator environment.is_uworker().
If a long-lived bot (is_uworker() == False) attempts to handle an untrusted testcase (trusted == False), it will trigger a fatal error (log_fatal_and_exit), stopping the processing immediately.

Testing

• Unit tests in minimize_task_test.py and test helpers in test_utils.py were updated to handle this new validation by default (setting trusted=True for test fixtures).
• Verified that all core tests pass.

Validation

[Insert logs showing a long-lived bot crashing when attempting to process an untrusted testcase]