Parallel passages extension

[ryderwishart]

We should leverage the copilot RAG index to reveal similar passages both in the translation draft and in the source text

• [x] Use two tabs (panels from web view ui toolkit), one for parallel passages as we have them now, and a second for Anomaly Detection
• [x] the Anomaly Detection tab reveals two lists of vrefs side-by-side (use 'flex' css for responsiveness), one under the "Your Draft" heading and another under "Source Text"
• [ ] show three kinds of mismatches: order mismatch, missing verses or extra verses
• [ ] use the combined greek/hebrew vref bible file for source text embeddings
• [x] if a verse hasn't been translated in the draft yet, then just grey it out from the results

Once we have this basic display working, let's try the following:

• [x] Detect anomalies in the editor itself using the language server

[dadukhankevin]

A couple of things I'm thinking about.

• I'm not sure if embeddings will be accurate enough for order mismatch to matter. We'll have to see.
• Does having this specific LAD on in the background all the time require too many resources? Switching databases, querying embeddings etc...

[ryderwishart]

A couple of things I'm thinking about.

• I'm not sure if embeddings will be accurate enough for order mismatch to matter. We'll have to see.
• Does having this specific LAD on in the background all the time require too many resources? Switching databases, querying embeddings etc...

Yeah we need some kind of benchmark with 'accepted' translations basically.

As for resources, I suspect metrics other than semantic similarity will be quicker. I was thinking TF-IDF might be faster and perhaps even better for what we want. Eventually we want an ensemble of metrics that can be indexed in the background, ideally a lot of more traditional algorithmic ones.

[ryderwishart]

The most pressing need for this feature is to improve the UI. Currently there are overflow problems, spacing issues (results take up lots of space), etc.

See #54