Review of plot_column_values() in draw.py

No paths hardcoded, make them a variable (e.g.1). We would like to be able to run this code for any file in any folder.
Include docstrings and specify types. Explain the purpose of each parameter. Here it's not clear to me what keep_test_mouse is for.

Example:

def add(a: int, b: int) -> int:
    """Add two integers.

    Args:
        a (int): First integer.
        b (int): Second integer.

    Returns:
        int: The sum of `a` and `b`.
    """

Here 'nothing' should become a global variable, you can include it before all the functions after the import section:

FILLNA_BEHAVIOR = 'nothing'
..
phase_data['manual_annot'] = phase_data['manual_annot'].fillna(FILLNA_BEHAVIOR)

If I got this correctly, this function does 2 things: 1) fillna in the manual annotation column 2) plots x and y position over time. Can we split this function into 2 functions? And can we use 'draw.py' only for the plotting functions and create another file for data cleaning / data preprocessing? It would be cool to start thinking of a "pipeline"
Again on this: it's not clear to me why we need to create subset of the whole dataset (e.g. data = data[cols] and phase_data = data[data['phase'] == phase]. Can you avoid it and just select the columns you need?
I would review the other functions below in terms of:

docstrings, types and description of variables and function
keep function short and constrained to one single step, when possible
avoid hardcoding parameters