fix: add serde default for truncation direction

[drbh]

This PR adjusts the similarity parameters to use a default for truncation direction

example request

curl "http://localhost:3000/similarity" \
-X POST \
-H "Accept: application/json" \
-H "Authorization: Bearer hf_XXXXX" \
-H "Content-Type: application/json" \
-d '{
    "inputs": {
        "sentences": [
            "That is a happy dog",
            "That is a very happy person",
            "Today is a sunny day"
        ],
        "source_sentence": "That is a happy person"
    },
    "parameters": {}
}'
# [0.85468096,0.9658465,0.68776166]

[Narsil]

LGTM.

The default is indeed right

https://github.com/huggingface/tokenizers/blob/14a07b06e4a8bd8f80d884419ae4630f5a3d8098/tokenizers/src/utils/truncation.rs#L7