Reduce allocations in PropertyDictionary

[Erarndt]

Fixes #

Context

Changes Made

Testing

Notes

[Copilot]

Pull Request Overview

This PR optimizes the PropertyDictionary by reducing allocations during property filtering. Key changes include preallocating the result list with the known collection count and using a struct enumerator when possible for performance benefits.

[Copilot]

[nitpick] Consider using 'propertiesCollection' for the type check (i.e. 'if (propertiesCollection is RetrievableValuedEntryHashSet hashSet)') to avoid redundant casting of _properties.

Suggested change

	ICollection<T> propertiesCollection = (ICollection<T>)_properties;
	List<TResult> result = new(propertiesCollection.Count);
	// PERF: Prefer using struct enumerators from the concrete types to avoid allocations.
	// RetrievableValuedEntryHashSet implements a struct enumerator.
	if (_properties is RetrievableValuedEntryHashSet<T> hashSet)
	{
	if (_properties is RetrievableValuedEntryHashSet<T> hashSet)
	{
	List<TResult> result = new(hashSet.Count);
	// PERF: Prefer using struct enumerators from the concrete types to avoid allocations.
	// RetrievableValuedEntryHashSet implements a struct enumerator.

[SimaTian]

This is weird or incomplete I think.
Is there any merit to the idea though?

[SimaTian]

Looks good, just please take a look at the Copilot suggestion if it makes sense. If not, just dismiss and we can move on.

[rainersigwald]

Is this branch used in practice? Should we add a Debug.Assert or something that it not become hot again?

[Erarndt]

I see one path where this would be used. There's one path that passes a ImmutableValuedElementCollectionConverter to the constructor which would hit the "else" path. I can check on how much that happens.