Introduced a new BatchedExecutor

[martindevans]

Created a new BatchedExecutor which processes multiple "Conversations" in one single inference batch. This is faster, even when the conversations are unrelated, and is much faster if the conversations share some overlap (e.g. a common system prompt prefix).

Conversations can be "forked", to create a copy of a conversation at a given point. This allows e.g. prompting a conversation with a system prefix just once and then forking it again and again for each individual conversation. Conversations can also be "rewound" to an earlier state.

Added two new examples, demonstrating forking and rewinding.

This is currently much "lower level" than the existing executors, and is really just a minimum viable system to move LLamaSharp over to batching. There is more work needed in the future:

• Wrap Conversation with functionality that the existing executors have (e.g. prompt templates, sampling)
• Extend Conversation with new capabilities (whatever is needed to handle out-of-context)
• Extend executor with new capabilities, such as saving/loading the entire batch.

How To Use

Brief guide to using the BatchedExecutor:

1. Create a BatchedExecutor
2. Create one or more new Conversations with executor.Prompt("hello");
3. Call executor.Infer() to run inference for all conversations which need inference simultaneously
4. Sample each conversation individually (conversation.Sample())
5. Prompt the conversation with the token chosen by sampling, or with more user input
6. goto 3

Conversation objects have flags which indicate what state they're in (waiting for sampling, waiting for inference) and will throw exceptions if you try to use them wrong.