fix: sanitize input text to remove null bytes and control characters

[echobt]

Description

This PR addresses an issue where null bytes (\x00) and other control characters in the input text cause the tokenizer to fail with a generic error.

Changes

• Added a sanitize_text function in src/core/embeddings.rs that filters out control characters while preserving newlines (\n), tabs (\t), and carriage returns (\r).
• Updated embed_batch to sanitize input text before passing it to the tokenizer.
• Added unit tests to verify the sanitization logic.

Testing

• Ran cargo test to verify the new sanitization logic and ensure no regressions.
• The new test test_sanitize_text passes, confirming that null bytes and unwanted control characters are removed.