[FEATURE] Improve error messages around schema

[kvnkho]

Is your feature request related to a problem? Please describe. When creating a transformer, the error message for invalid annotations is currently very vague. The following example:

from typing import List
from fugue import transform
import pandas as pd

def map_phone_to_location(df: List[List]) -> List[List]:
    for row in df:
        row.append(["test"])
    return df

data = pd.DataFrame({"phone": ["(217)-123-4567", "(217)-234-5678", "(407)-123-4567", 
                               "(407)-234-5678", "(510)-123-4567"]})
transform(data, map_phone_to_location, schema="phone:str, new_col:str")

produces an error like:

FugueInterfacelessError: ('<function map_phone_to_location at 0x7fd1894c5b00> is not a valid transformer', TypeError("Input types not valid IndexedOrderedDict([('df', [Other])]) for <function map_phone_to_location at 0x7fd1894c5b00>"))

Describe the solution you'd like

Instead, I think this error should say something like:

"Detected type "List[List]". The valid type annotations are: ...."

[kvnkho]

To add to this, missing type annotations also give an unhelpful error:

def clip(df: pd.DataFrame):
    df['value'] = df['value'].clip(1,2)
    return df

Gives:

FugueInterfacelessError: ('<function clip at 0x7fe34365c710> is not a valid transformer', FugueInterfacelessError("* can't be used on cotransformer output schema"))