Add ensemble_batch_size for single-device inference

[randommm]

Issue

#905

Motivation and Context

On devices with large amounts of RAM like Strix Halo, this can greatly speed up results

Public API Changes

• No Public API changes
• Yes, Public API changes (Details below)

How Has This Been Tested?

local testing

Checklist

• The changes have been tested locally.
• Documentation has been updated (if the public API or usage changes).
• A changelog entry has been added (see changelog/README.md), or "no changelog needed" label requested.
• The code follows the project's style guidelines.
• I have considered the impact of these changes on the public API.

[chatgpt-codex-connector]

Codex usage limits have been reached for code reviews. Please check with the admins of this repo to increase the limits by adding credits.
Credits must be used to enable repository wide code reviews.

[gemini-code-assist]

Code Review

This pull request implements ensemble batching for single-device predictions in TabPFN, enabling multiple compatible ensemble members to be processed in a single forward pass to improve performance. The changes introduce an ensemble_batch_size parameter across the API and update the inference engines to handle batched outputs. Feedback focuses on fixing a potential shape mismatch in embedding extraction, addressing an unused parameter in the on-demand engine, and standardizing telemetry timing for model execution.

[adrian-prior]

Hey @randommm,

Thanks for setting this up, and apologies for taking so long to get back on this! I do have some open questions, which I left on the PR, and there are also a bunch of comments left by Gemini, of which I think many are valid. Would you mind going through them?

[randommm]

@adrian-prior i think it should be all fixed now.