Fix TestTurbiniaTaskBase to instantiate the requested evidence class

[rootkiller6788]

Summary

TestTurbiniaTaskBase.setUp() accepts an evidence_class parameter and stores it on the instance, but then unconditionally instantiates self.evidence as evidence.RawDisk. As a result, any test that passes a different evidence class silently runs against the wrong evidence type.

For example, VolatilityTaskTest passes evidence_class=RawMemory and then sets self.evidence.profile / self.evidence.module_list, but the object is actually a RawDisk.

Change

Instantiate the requested class instead of hardcoding RawDisk:

self.evidence = self.evidence_class(source_path=test_disk_path)

All evidence classes inherit from Evidence and accept source_path via **kwargs, so this is safe for every existing caller (RawMemory, ReportText, BodyFile, DiskPartition, DockerContainer, ContainerdContainer, BulkExtractorOutput, BinaryExtraction, MachoExtraction, ElfExtraction, PhotorecOutput).

Verification

• Confirmed against HEAD that evidence_class was stored (workers_test.py:47) but never used, while line 69 hardcoded evidence.RawDisk.
• Grepped all evidence_class= callers in the test suite to confirm the parameter is genuinely exercised.
• python -m py_compile passes on the modified file.

Fixes #1373

[aarontp]

lgtm, running tests to make sure they pass.

[rootkiller6788]

Thank you for the review and for kicking off the test run! The CI failure it surfaced was caused by the base-class change itself (issue #1373): TestTurbiniaTaskBase now honors evidence_class, and two test files were passing their output evidence classes as the task's input evidence class. partitions_test passed DiskPartition (whose .name raises TypeError because partition_location is None) and photorec_test passed PhotorecOutput (which has no device_path attribute that PhotorecTask.run accesses), producing the 6 test errors. Since both tasks consume a raw disk image as input, I've updated those tests to use evidence.RawDisk (the base default), which restores the previous behavior for these tests while preserving the #1373 fix for any test that supplies a real custom evidence class. I verified all 6 affected tests pass locally. Could you please take another look / re-run the test workflow?