feat: clean up working dir after each run

[IvanJobs]

Task name:
Clean up working dir after each run to avoid permission issue.

Description:
Oryx build will change file owners to root after each build which caused permission issue on AzDo with self-hosted agents when Cx tries to checkout the code by the second time during which it will fail to delete the previous files owned by root.

https://github.com/microsoft/Oryx/blob/0b76566110f0db32097b869761e056fd9b01848d/src/BuildScriptGenerator/Node/NodeBashBuildSnippet.sh.tpl#L142

This PR is to clean up files on behalf of root inside the docker container to prepare a clean working dir for next run.

The issue can be described like this.

When cx uses self hosted agent, the source code will be checked out into some folder on the agent, like work/1/s, I think s stands for source code. The file owner is the current user by which we start the self-hosted agent, for example, ivan. A docker container will be created with work/1/s mounted to /working_dir in the container. When oryx build finished build the project in container, it will change file owners to root for all files under /working_dir, i.e., _work/1/s on the hosted machine. The second time, when azdo pipeline tries to check out the new version of the repo, it will try to clean up _work/1/s on behalf of ivan, but it's unable to delete files owned by root.

based on my testing, anyway in this case, self-hosted agent will try to clean up or overwrite this folder _work/1/s before each run. It should be good for us to clean it up after each run in the container.

Test it from end-to-end in my private AzDo repo, looks good.

Customized version:

Changes applied:

Run multiple times, looks good.

Documentation changes required: (Y/N)
N

Added unit tests: (Y/N)
N/

Attached related issue: (Y/N)
N

Checklist:

• Task version was bumped - please check instruction how to do it
• Checked that applied changes work as expected
• Reproduced the permission issue for the second run
• Change the launch script and try to reproduce it, failed with successful results
• Try managed agent, things work good