Defer Blackbox Fuzzer Log Uploads

[IvanBM18]

WIP PR

Bug: b/532610906

Context

• This PR comes as a need after the revert of this Revert "Fix legacy Datastore access in data_handler.py for UWORKERs" #5339 PR which introduced short circuit mechanisms to avoid DB queries, this however proved to provoke issues, so short circuit is no longer an option to fix any PERMISION_DENIED erros.
• There were 2 instances of PERMISSION_DENIED issues, the first one is already solved trough this [NDB PERMISSION_DENIED error] Fix trusty stacktrace on ANDROID_EMULATOR #5375

Overview

This is the last PR to tackle an instance of PERMISSION_DENIED issues when executing a BlackBox test case in a uworker bot.

This PR refactors uses the utils, classes and refactors to previous PRs to defer the blackbox fuzzer log upload to the posprocess step. Since the blackbox fuzzer logs can be larger than regular Coverage guided engine fuzzer logs, we cant just simply save its content into the crash queue, instead we save the fuzzer output into a temp file and we save that file in the crashes queue.

Then when we finish executing all the test cases(which can be multiple concurrent ones), we drain the crash queue, retrieving the log file paths and transforming them into an istance of FuzzerRunOutputData class, which is then placed into the fuzzer ssesion output to be handled & uploaded at the postprocess step

Changes

• Fuzzer log for BB fuzzer are now saved into temp files instead of the actual string
  • This is done because we use a special type of queue to handle multiple concurrent fuzzing outputs, this queue has a hard limit on the amount of data that it can handle,
    • So if a single fuzzer uploads to much info then the queue will block operations until someone pulls data from the queue
    • if we have a job that runs multiple test cases simultaneously & a fuzzer that logs too much info we can end up in a livelock multithreading problem
• Creates a new method to drain the queue for multiple crash results
  • Encapsulates the fuzzer logs from the queue using the FuzzerRunOutputData
• Since we no longer upload the logs at the utask_main portion, we deleted the upload parameter for the run_tetcase method

Tests performed

Changes in dev since July 23rd, from which we can see how the errors dramastically went down after the changes were introduced:

The other 403 issues that still appear in dev are, from custom_binary tests being performed by teammates in dev, if we remove those logs, we can see how the errors go to 0

PR stack

PR #5376: Introduce FuzzerRunOutputData and utils (Base: master)
PR #5377: Refactor _to_fuzzer_run_output to use FuzzerRunOutputData (Base: PR 1)
PR #5388: refactor: Add utils.create_temp_file helper (Base: PR 2)
Current -> PR #5378: Defer Blackbox Fuzzer Log Uploads (Base: PR 3)