Correctly infer bool and object types in autoML

[arnavgarg1]

Fixes an issue where is_field_boolean always return True if there are greater than 3 distinct values. This should return False by default if there are greater than 3 values.

Additionally, some of the type checking fails because source.get_distinct_values() actually gets unique values after dropping NaNs. This prevents us from catching the case we're looking for, i.e., 3 distinct values of which one is a None/NaN. This logic is modified to check for None and Nans, and return True if either of those are found in the case that there are 3 distinct values.

Co-authored-by: @jppgks

[github-actions]

Unit Test Results

         6 files  ±  0           6 suites  ±0   3h 56m 42s ⏱️ + 35m 41s
  3 527 tests +11    3 445 ✔️ +10    82 💤 +1  0 ❌ ±0 
10 581 runs  +33  10 317 ✔️ +30  264 💤 +3  0 ❌ ±0 

Results for commit 8c19ad1. ± Comparison against base commit 87ca887.

♻️ This comment has been updated with latest results.

[magdyksaleh]

I would have assumed that the fix would be

def is_field_boolean(source: DataSource, field: str) -> bool:
    num_unique_values, unique_values, _ = source.get_distinct_values(field, max_values_to_return=4)
    if num_unique_values <= 3:
        for entry in unique_values:
            try:
                if np.isnan(entry):
                    continue
            except TypeError:
                # For some field types such as object arrays np.isnan throws a TypeError
                # we catch it since we know in this case it is not a bool.
                return False
            if isinstance(entry, bool):
                continue
            return False
        return True   
    return False

[arnavgarg1]

I would have assumed that the fix would be

def is_field_boolean(source: DataSource, field: str) -> bool:
    num_unique_values, unique_values, _ = source.get_distinct_values(field, max_values_to_return=4)
    if num_unique_values <= 3:
        for entry in unique_values:
            try:
                if np.isnan(entry):
                    continue
            except TypeError:
                # For some field types such as object arrays np.isnan throws a TypeError
                # we catch it since we know in this case it is not a bool.
                return False
            if isinstance(entry, bool):
                continue
            return False
        return True   
    return False

So this is what I initially did, but the problem is that the unique_values returned from source.get_distinct_values always drops the NaNs before returning the unique values. So if you have a column with 3 distinct values ['a', 'b', np.nan], we're only going to get 2 values back. The result of that is that np.isnan() will always fail and it'll return False in the except block, but IMO, this is actually a valid bool type field.