Optimization - A queue with fixed storage size backed by a circular buffer

[SignalRT]

This should provide some optimizations to the FixedQueue. My test are:

BEFORE THE CHANGES:

AFTER THE CHANGES:

[Copilot]

Pull Request Overview

This PR optimizes the FixedQueue implementation by replacing the naive List-backed approach with a circular buffer, along with several other performance improvements throughout the codebase. The changes focus on reducing memory allocations and improving computational efficiency.

• Reimplemented FixedSizeQueue using a circular buffer instead of List with Remove operations
• Optimized string concatenation in StreamingTokenDecoder using spans where available
• Replaced LINQ operations with more efficient alternatives in multiple executors and BatchedExecutor

Reviewed Changes

Copilot reviewed 8 out of 8 changed files in this pull request and generated 4 comments.

Show a summary per file

File	Description
LLama/Common/FixedSizeQueue.cs	Complete rewrite using circular buffer architecture with optimized enqueue/dequeue operations
LLama/StreamingTokenDecoder.cs	Optimized string creation using CollectionsMarshal.AsSpan for .NET 5+
LLama/LLamaExecutorBase.cs	Added AntipromptProcessor integration and optimized session token management
LLama/LLamaInteractExecutor.cs	Replaced token-based antiprompt checking with string-based AntipromptProcessor
LLama/LLamaInstructExecutor.cs	Replaced token-based antiprompt checking with string-based AntipromptProcessor
LLama/Batched/BatchedExecutor.cs	Optimized batch queue management using head pointer instead of RemoveAt operations
LLama/AntipromptProcessor.cs	Minor variable rename for clarity
LLama.Benchmark/Collections/FixedSizeQueueBenchmark.cs	Added benchmark tests for the new FixedSizeQueue implementation

---

_{Tip: Customize your code reviews with copilot-instructions.md. Create the file or learn how to get started.}

[martindevans]

LGTM, just a couple of minor nits