[Fiducial] Changes to angle variables due to possible Awkward 1 bug and fixed merge_root.py bug

The implementation of the new angle variables to the fiducial processor introduced in MR !498 (merged) were throwing an error when testing with Coffea 2025 and also drastically reduced the number of events when running over all samples.

We went back to a slightly improved version of the original implementation (see for example here). A small validation was done, showing that the distributions were looking very similar.

Closes #280 (closed)