Define expected return type for ungrouped `count()`

Problem

df.count() currently behaves like df.rowsCount() for ungrouped DataFrames.

This creates an inconsistency:

• ungrouped df.count() returns a scalar value
• grouped df.groupBy { ... }.count() returns a DataFrame

For users, this breaks the expectation that count() is a DataFrame operation with consistent input/output shape.

Design question

Should count() always return a DataFrame?

Options:

• keep current behavior
• change ungrouped count() to return a DataFrame
• introduce a new API and deprecate the inconsistent behavior

Expected

Define the intended contract of count() before 1.0.

Acceptance criteria

• Decision made on count() return type for ungrouped DataFrames
• Behavior is implemented or explicitly documented
• rowsCount() and count() responsibilities are clearly separated
• Tests cover both grouped and ungrouped count()

Motivation

count() is a core aggregation API.

Its return type is part of the public API contract. If we want to change it, this must be decided before 1.0 to avoid a breaking change later.

Currently df.rowsCount() and df.count() do the same ie are redudnant for ungrouped data-frames. However, for sake of consistency, count() should always return a data-frame irrespective if data is grouped or not. 

In particular, this type of consistency made dplyr hugely popular, because it guaranteed that all major functions work df-input -> df-output without exception (grouping or not)