pandas: fix various arguments for Series.iloc and DataFrame.loc

[Dr-Irv]

pylance 2021.7.3

The following code:

import pandas as pd

df = pd.DataFrame({"x": [1, 2, 1], "y": [3, 4, 5]})


foo = df.loc[:, "x":"y"]

reveal_type(foo)


s = pd.Series([1, 2, 3])

soo = s.iloc[1:2]

s2 = pd.Series([1, 2, 3], index=["a", "b", "c"])

i = pd.Index(["a", "c"])

ns = s2.loc[i]

generates the following messages with Python 3.8:

Type of "foo" is "str | bytes | date | datetime | timedelta | bool | int | float | complex"

No overloads for "__getitem__" match the provided arguments
  Argument types: (slice)

No overloads for "__getitem__" match the provided arguments
  Argument types: (Index[str])

With python 3.7, the first message is correct:

Type of "foo" is "DataFrame"

I don't fully understand why the python version mattered, and why with python 3.8, it was matching a different signature, but this PR fixes all 3 issues for the .loc for DataFrame and .iloc for Series and .loc for Series