[dev] Make the gRPC client init test more resilient on Windows

[kittaakos]

Describe the problem

#2023 added integration tests to check whether the CLI daemon can recover from an inconsistent directories.data state during the core client initialization.

The tests do the followings:

• start the CLI daemon,
• initialize a valid state,
• stop the CLI daemon,
• corrupt the directories.data folder by deleting platform tools, primary package, and library indexes,
• start the CLI daemon,
• run init,
• verify if the daemon is functional.

The tests tend to fail on Windows. As also noted in #2028, deleting the platform tools might fail because the OS is holding the process.

The tests must do the following to be more resilient:

• initialize valid directories.data structure via the CLI using ./arduino-cli core update-index,
• save the valid directories.data folder,
• each test case should copy the correct folder and corrupt it.

To reproduce

Please review the issue description and check for flaky builds on the CI.

Example:

 1) core-client-provider
       should recover when the 'packages\builtin\tools\serial-discovery' folder is missing:
     Error: EPERM: operation not permitted, unlink 'c:\Users\RUNNER~1\AppData\Local\Temp\d-2023410-708-166j91u.ub7p\.testArduinoIDE\data\Arduino15\packages\builtin\tools\serial-discovery\1.4.0\serial-discovery.exe'
      at Object.unlinkSync (node:fs:1722:3)
      at fixWinEPERMSync (D:\a\arduino-ide\arduino-ide\node_modules\rimraf\rimraf.js:219:13)
      at rimrafSync (D:\a\arduino-ide\arduino-ide\node_modules\rimraf\rimraf.js:319:28)
      at D:\a\arduino-ide\arduino-ide\node_modules\rimraf\rimraf.js:350:5
      at Array.forEach (<anonymous>)
      at rmkidsSync (D:\a\arduino-ide\arduino-ide\node_modules\rimraf\rimraf.js:349:26)
      at rmdirSync (D:\a\arduino-ide\arduino-ide\node_modules\rimraf\rimraf.js:342:7)
      at fixWinEPERMSync (D:\a\arduino-ide\arduino-ide\node_modules\rimraf\rimraf.js:217:5)
      at rimrafSync (D:\a\arduino-ide\arduino-ide\node_modules\rimraf\rimraf.js:319:28)
      at D:\a\arduino-ide\arduino-ide\node_modules\rimraf\rimraf.js:350:5
      at Array.forEach (<anonymous>)
      at rmkidsSync (D:\a\arduino-ide\arduino-ide\node_modules\rimraf\rimraf.js:349:26)
      at rmdirSync (D:\a\arduino-ide\arduino-ide\node_modules\rimraf\rimraf.js:342:7)
      at fixWinEPERMSync (D:\a\arduino-ide\arduino-ide\node_modules\rimraf\rimraf.js:217:5)
      at rimrafSync (D:\a\arduino-ide\arduino-ide\node_modules\rimraf\rimraf.js:319:28)
      at Context.<anonymous> (lib\test\node\core-client-provider.slow-test.js:86:27)
      at processTicksAndRejections (node:internal/process/task_queues:96:5)

Expected behavior

The CI is not failing randomly.

Arduino IDE version

36e2092

Operating system

Windows

Operating system version

N/A

Additional context

When the CI is red due to failed tests, GH Actions will report a false positive defect with the certificates:

The job was canceled because "windows-2019_WINDOWS_SIGNING_CERTIFICATE_PFX_WINDOWS_SIGNING_CERTIFICATE_PASSWORD_pfx" failed.

Issue checklist

• I searched for previous reports in the issue tracker
• I verified the problem still occurs when using the latest nightly build
• My report contains all necessary details

[kittaakos]

We have to fix this ASAP, as the nightly builds are flaky.