Write a file of embeddings to Parquet from eval recs generation

[karlhigley]

This grabs the candidate embeddings from each pipeline execution and builds up a cache of embeddings over the course of the eval run. Once the eval run finishes, it writes them to a two column Parquet file where the first column is the article id and the second is the corresponding embedding.

[sophiasun0515]

lgtm!

[mdekstrand]

This looks good to me for doing the saving, but will leave the code in an incompatible state for doing the measurements and automating the run — up to you if you want to fix that in this PR or a second PR.

I like the design of specifying an output directory, this will give us future flexibility to add more output files.

Things that are needed to finish wiring into measurement:

1. dvc.yaml — change the recommend- stages to specify an output directory, and correct the dependency and output files.
2. update measure.py to take an input directory, and look for recommendations.parquet in it.
3. dvc.yaml — update the measure- stages as well.

[mdekstrand]

Need to add the embeddings.parquet files to outs in dvc.yaml (so DVC will track / push / pull them), but otherwise looks good. One further comment on a now vs. later consistency refactor.