Exercise: median per group

Write a function group_medians(df, by, column): it groups the rows of the table df by the column by and returns a dictionary. A key of the dictionary is a value from by, and its value is the median of the column column in that group.

• df = pd.DataFrame({"team": ["a", "a", "a", "b"], "salary": [100, 400, 200, 50]}), group_medians(df, "team", "salary") → {"a": 200.0, "b": 50.0}
• the same teams, "salary": [100, np.nan, 120, 80] → {"a": 110.0, "b": 80.0}
• "team": ["a", "a", "b", "b"], "salary": [100, 120, np.nan, np.nan] → {"a": 110.0}

Constraints: the column by has no missing values, and column holds numbers and missing values; with an even number of values the median is the mean of the two central ones; missing values are skipped, and a group without a single value is left out of the dictionary.

Do not add extra imports unless the task says so explicitly.