[integration] Add bm25s library for bm25 retrieval models/indices (…

…#763)

Hello!

## Pull Request overview
* Add BM25S as a library on the Hub. 

## Links
* Repository: https://github.com/xhluca/bm25s
* Models: https://huggingface.co/models?other=bm25s

## Details
BM25S is a very new library for efficient BM25: an important algorithm
for full-text search. It can be combined with vector search (e.g.
Sentence Transformers' domain) for hybrid search, which is commonly used
and very powerful. I'm considering mentioning this library more in
Sentence Transformers and its example/docs, so it'll be nice to get the
"Use in Library" button & download count for these models/indices. I've
set `filter` to False as it's not a big library currently.

cc @xhluca for context: this should add a "Use this model" button in the
top right of all models on the Hub that have a `bm25s` tag in the model
card metadata. It'll then show the following snippet:
```python
from bm25s.hf import BM25HF

retriever = BM25HF.load_from_hub("${model.id}")
```

- Tom Aarsen

packages/tasks/src/model-libraries-snippets.ts

@@ -56,6 +56,12 @@ export const bertopic = (model: ModelData): string[] => [
 model = BERTopic.load("${model.id}")`,
 ];
 
+export const bm25s = (model: ModelData): string[] => [
+	`from bm25s.hf import BM25HF
+
+retriever = BM25HF.load_from_hub("${model.id}")`,
+];
+
 const diffusers_default = (model: ModelData) => [
 	`from diffusers import DiffusionPipeline
 

packages/tasks/src/model-libraries.ts

@@ -111,6 +111,16 @@ export const MODEL_LIBRARIES_UI_ELEMENTS = {
 			wildcard: { path: "*.npz" },
 		},
 	},
+	bm25s: {
+		prettyLabel: "BM25S",
+		repoName: "bm25s",
+		repoUrl: "https://github.com/xhluca/bm25s",
+		snippets: snippets.bm25s,
+		filter: false,
+		countDownloads: {
+			term: { path: "params.index.json" },
+		},
+	},
 	chat_tts: {
 		prettyLabel: "ChatTTS",
 		repoName: "ChatTTS",