Add additional generic types to DataFrame methods

[controversial]

Adding generic types to a few more methods beyond what was added in #293 by @scarf005

Focusing mostly on adding identity types to methods which I believe don’t change the original type of the dataframe. I added “identity” type signatures to the following methods:

extend, fillNull, filter, interpolate, limit, max, mean, median, min, quantile, rechunk, shiftAndFill, shrinkToFit, slice, sort, std, sum, tail, unique, var, vstack, where, upsample

These previously returned DataFrame<any>, even when called on a well-typed DataFrame, but now return DataFrame<T> (the original type)

---

I also added better types for a few slightly more complex ones:

• map
  • improved return type based on the function passed, but unimproved parameter type
• nullCount
• toRecords
• toSeries
  • for now, returning a broad union type, rather than identifying the specific column by index
• withColumn

---

Along the way, I added minor fixes for the types of:

1. pl.intRange [1] which had overloads in the wrong order leading to incorrect return types, and
2. the pl.Series(name, values, dtype) constructor [2], whose strongly-typed overload was failing to apply in simple cases like pl.Series("index", [0, 1, 2, 3, 4], pl.Int64) when the input array used numbers instead of BigInts

[controversial]

will work on test failures

[controversial]

Have now addressed the previous test failures