fix(import): support for nested namespace packages

[kirat-singh]

If multiple directories in sys.path provide a nested namespace package, then jedi would only visit the first directory which
contained the package. Fix this by saving the remaining path list in the ImplicitNamespaceValue and add a test for it.

Added test namespace packages to demonstrate the failure:

namespace1/ namespace1/pkg1 namespace1/pkg1/pkg2 namespace1/pkg1/pkg2/mod1.py namespace2 namespace2/pkg1 namespace2/pkg1/pkg2 namespace2/pkg1/pkg2/mod2.py

If both namespace1 and namespace2 are added to sys path, then jedi sees only mod1 without the fix. After the fix it correctly sees both mod1 and mod2.

[kirat-singh]

Whoops this is a duplicate of #1784 I'm happy to withdraw this if the other one is approved.

[codecov-commenter]

Codecov Report

Merging #1805 (53e8370) into master (65bc1c1) will not change coverage.
The diff coverage is 100.00%.

@@           Coverage Diff           @@
##           master    #1805   +/-   ##
=======================================
  Coverage   94.50%   94.50%           
=======================================
  Files          80       80           
  Lines       11817    11817           
=======================================
  Hits        11168    11168           
  Misses        649      649           

Impacted Files	Coverage Δ
jedi/inference/imports.py	98.62% <100.00%> (ø)

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 65bc1c1...53e8370. Read the comment docs.

[davidhalter]

I guess the isinstance there is kind of weird and does not make sense. I guess just change it to

        file_io_or_ns, is_pkg = inference_state.compiled_subprocess.get_module_info(
            string=import_names[-1],
            path=paths,
            full_name=module_name,
            is_global_search=False,
        )
        if is_pkg is None:
            return NO_VALUES

The for loop here is removed and not needed anymore. I do not know why it was there in the first place. Probably some refactorings that made it unnecessary.

[davidhalter]

#1784 is a duplicate, but it changes some other things as well, so we might as well merge both and resolve the merge conflict :)

The earlier fix will probably just win ;-).

[kirat-singh]

thanks for looking, and agree should have simplified. changed it as you suggested.

[davidhalter]

Thanks!