[13.x] Fix failed job providers crashing on corrupted payloads

[bipinks]

Closes #59635.

Problem

The three UUID-based failed job providers call json_decode($payload, true)['uuid'] with no guard. If the queue backend ever delivers a corrupted / non-JSON payload (Redis network corruption, truncated SQS body, custom driver bug), json_decode returns null, null['uuid'] throws a TypeError on PHP 8+, the exception propagates out of Worker::logFailedJob, and the failed job record is never written.

The safety net for jobs that fall through the cracks has a hole — the job disappears silently with no retry button and no audit trail.

Fix

Decode defensively in each provider:

$decoded = json_decode($payload, true);
$uuid = is_array($decoded) ? ($decoded['uuid'] ?? null) : null;

• FileFailedJobProvider — stores null as the id; the JSON file tolerates it.
• DatabaseUuidFailedJobProvider — generates a fallback Str::uuid() so the NOT NULL constraint on failed_jobs.uuid isn't violated.
• DynamoDbFailedJobProvider — same fallback, since DynamoDB 'S' attributes can't be null.

Tests

Added RED-verified tests for each provider covering three corruption shapes: unparseable JSON, JSON that decodes to a scalar, and JSON missing the uuid key. All 212 tests/Queue/* tests continue to pass.