CI: Fix Azure Pipelines problem with PATH

[datapythonista]

We should be able to set the environment variable PATH in an Azure Pipelines job with echo '##vso[task.prependpath]$HOME/miniconda3/bin'. See https://github.com/pandas-dev/pandas/blob/master/azure-pipelines.yml#L27

But that didn't work when we implemented it, so we ended up setting the path manually in every step (e.g. https://github.com/pandas-dev/pandas/blob/master/azure-pipelines.yml#L48).

This is adding noise to our CI configuration, making it harder to see what every step is doing. We should try to understand what's the problem with setting the PATH, and fix it.

This also applies to ci/azure/posix.yml (besides azure-pipelines.yml in the root).

[bhavaniravi]

I would like to work on this. Since I have gone through this part in the mentoring side, I feel confident and it is indeed an interesting problem to solve

[datapythonista]

Thanks a lot @bhavaniravi. Let me know if you need help.

[bhavaniravi]

Just one question for now, How do you test the changes in CI? I can think of setting an azure-pipeline for my forked repo? Is that a way to go?

[datapythonista]

I think that's quite smart. It took me couple of months to realize, waiting for the pandas CI to run jobs (some times there is a long queue in the CI). There is something weird, because in my fork's CI some things fail that don't fail in pandas, and that shouldn't happen. But regardless of that, I think it's the best approach if you plan to do significant work in the CI.