Preserve element order when initializing Python dictionary

[philipturner]

Pandas DataFrames were rendering incorrectly in IPython. This bug was reported in philipturner/swift-colab#22, and seems to have been around in the swift-jupyter era. The culprit is how a dictionary literal currently converts into a Python object.

PythonKit/PythonKit/Python.swift

Lines 1466 to 1468 in f12defa

	public init(dictionaryLiteral elements: (PythonObject, PythonObject)...) {
	self.init(Dictionary(elements, uniquingKeysWith: { lhs, _ in lhs }))
	}

The new implementation preserves element order in the final Python object, unlike Dictionary.pythonObject. When keys are duplicated, throw the same runtime error as Swift.Dictionary.init(dictionaryLiteral:). This differs from Python's key uniquing semantics, which silently override an existing key with the next one it encounters.

The previous implementation used Dictionary.init(uniquingKeysWith:) and retained only the first duplicate key's value. Changing the rule introduces API breakage, but is the most semantically correct. We want to create a Swift-first API, and duplicating keys in a literal is bad/unsafe practice.

Additional changes

• Fixed malformatted comment above prefix func -.
• Adjusted Members initializer of PythonClass to follow the uniqueness semantics above. The alternative initializer uses Dictionary, which up until now had different behavior upon key duplication.

[philipturner]

I also have a working private notebook where I modified the PythonKit embedded into Swift-Colab, and Pandas DataFrames rendered correctly.