✨ Add MNIST simulator

[taxe10]

This PR introduces an alternative simulator that uses the MNIST dataset to generate digit-based test data. The simulator downloads MNIST, ingests a subset of the images into Tiled under:

f"{RESULTS_TILED_URI}/mnist"

Ten containers are created—one for each digit (0–9)—with a predefined number of images per digit. Container sizes vary, and a few contain only a single image, similar to the beamline 733 use case.

Each streamed message includes a simple feature vector of the form:

[digit, index]

where:

• digit is the numeric label (0–9) corresponding to the container,
• index is the position of the image within that digit’s subset.

This enables a simple and reproducible source of data for testing vector streaming and feature processing.

Usage

To enable the simulator, update your configuration:

arroyo_vec_sim:
  ...
  environment:
    SIMULATION_TYPE: "mnist"  # Use "mnist" for MNIST simulation

[xiaoyachong]

Hi Tanny, thanks for your PR. I noticed a callback error when selecting an image with the digit 3 or 5, where the container contains only one image.