Improved trajectory_projector calculations through vectorization

[awestphal1]

Closes #518 #519 #520 #521

The trajectory_projector functions now have significantly better performance due to vectorization. Furthermore, I replaced large parts of the calculations with shapely and numpy methods. The functions now work directly on Shapely geometries, so a geo_data structure is no longer necessary.

The correct_invalid_trajectory function now also returns a list of all person IDs containing invalid trajectory points (similar to the outlier detection). This makes it possible to plot the modified trajectories before and after correction for easier comparison. I also added corresponding code to the preprocessing notebook.

[schroedtert]

Overall looks great, have only looked at the latest three commits containing the changes for the vectorization improvement.

It would be great if you could move the shapely normalization to the constructor of the WalkableArea (#521 and see comment). And if you could take a look at the missed lines in the code coverage and construct test cases that would cover them.

[JuleAdrian]

Looks good overall! The method does not appear in the API references.
I think we should make an entry in methods.rst.

[codecov]

Codecov Report

❌ Patch coverage is 98.66071% with 6 lines in your changes missing coverage. Please review.
✅ Project coverage is 92.01%. Comparing base (24d747a) to head (4a5b106).

Files with missing lines	Patch %	Lines
...edpy/preprocessing/trajectory_outlier_detection.py	98.61%	3 Missing ⚠️
pedpy/preprocessing/trajectory_projector.py	98.70%	3 Missing ⚠️

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.