FIBSEM metadata visualization

[minnerbe]

This PR adds a script that visualizes metadata for a given render stack as a series of HTML plots together with a summary/landing page.

In the future, this is intended to run as an automated job for a dataset while imaging, storing the outputs in a directory that is accessible from the dataset's github issue. To test, I used the jrc_mus-heart-5 dataset, which currently has about 11k tiles. Using 16 processes, the whole process took about 1min and didn't use any significant amount of memory, so this can be run with a single 16-core job on our cluster.

Script usage

usage: aggregate_metadata.py [-h] --base-data-url BASE_DATA_URL --owner OWNER --project PROJECT --stack STACK [--output-dir OUTPUT_DIR] [--n-workers N_WORKERS]

Fetch all tile specs for a render stack and visualize the metadata.

optional arguments:
  -h, --help            show this help message and exit
  --base-data-url BASE_DATA_URL
                        Render web services host (e.g. em-services-1.int.janelia.org:8080).
  --owner OWNER         Render owner for the project.
  --project PROJECT     Render project name.
  --stack STACK         Render stack to query for tile specs.
  --output-dir OUTPUT_DIR
                        Directory where HTML files for plots will be written.
  --n-workers N_WORKERS
                        Number of worker processes to use for fetching tiles.

Pixi

Since I had problems with the previous way of setting up environments (a mix of conda and poetry), I moved the whole environment management to pixi. Installing and activating the environment is as easy as pixi shell in the directory containing the pyproject.toml file. The only difference to the previous environment is that the new environment also includes the janelia_emrp package as a python dependency, so no manual modification of the path is required anymore.

Instructions about how to use pixi are in the readme, I tested the setup on my laptop (mac/arm), my workstation, and the cluster (both linux/x86).