Sometimes no analysis for pip installed libraries

Some libraries installed via pip such as numpy end up with an message like:

[Info  - 12:17:31 PM] Import not found: numpy

And no analysis info.

Sometimes the analysis gets hung up on analyzing some pip related libraries and appears to not proceed, but that may be another issue.

To verify with the dependency handling. Current imports code is not bullet proof.

Also on a Django project:

Analysis queued: cool
[Info  - 1:37:08 PM] Import not found: django.db
Import:  django.db c:\Users\jabaile\Desktop\djangoo\venv\lib\site-packages\django\db\__init__.py 
Analysis queued: django.db
[Info  - 1:37:08 PM] Import not found: django.core
Import:  django.core c:\Users\jabaile\Desktop\djangoo\venv\lib\site-packages\django\core\__init__.py

I found where 'import not found' is coming from. It is red herring - typeshed resolution happens to report this way that the stub is not found. Fixed that with a better message. Secondary issue, which is, in fact, caused import to start looking for a stub is that file inside workspace was being considered library rather than user code.

There is recursive import in scikit-learn, opening _k_means eventually leads to spectral.py which again imports from .k_means_ import k_means. Current import/dependency resolution does not handle this reliably. Needs #432. However, I handled recursion via RDT since the module is already there - although its analysis is incomplete.