[KED-1258] Allow logging directory to be specified

Description

It is very jarring for kedro to change directories on its users. It appears that this is done for logging purposes. Is there a way to specify the logging directory without changing the users directory? This is possibly related to #149.

https://github.com/quantumblacklabs/kedro/blob/c96e94dc46948b05d9d09e285894ba3ff03c2595/kedro/context/context.py#L518-L522

Context

It is very confusing to have a side effect of changing directories after running a function. This behavior is not typical with other python frameworks/libraries that I am familiar with.

Possible Implementation

set the logging directory and remove the os.chdir line.

logging.basicConfig(filename=str(project_path / log_filename), filemode='w')

Possible Alternatives

My project templates track the user’s directory and changes it back to where the user intended to be. This works, but is less desirable.

[WaylonWalker]

Great Work!!! Excited to see this in the next version. We have a quick hack implemented on our pipelines that simply saves the current directory, loads context, then changes back to where the user was.

It might be good to update run.py from the template to not rely on the user’s current working directory.

https://github.com/quantumblacklabs/kedro/blob/develop/kedro/template/{{ cookiecutter.repo_name }}/src/{{ cookiecutter.python_package }}/run.py#L54

def run_package():
    # entry point for running pip-install projects
    # using `<project_package>` command
-   project_context = load_context(Path.cwd())
+   project_context = load_context(Path(__file__).parents[1]) # it might be a different layer of parents double check that
    project_context.run()

[limdauto]

Great shout. Will definite make that change.