fix: catch RuntimeError for missing torch import

chore: update pyproject

chore: fix numerical inconsistencies

test: add tests for tokenizers

chore: fit sparse encoders after route modifications in HybridRouter

fix: remove decorator around the HybridRouter.__init__

fix: give the functional BM25 a distinct filename

fix: fix route fitting for add/strategy

fix: _async_encode should take a sparse encoding function

fix: empty sparse embeddings

fix: _encode / _async_encode match superclass signature

- Deleted `pinecone_local.py` and `10-pinecone-local.ipynb`
- Updated `pinecone.py` to handle both cloud and local Pinecone index scenarios
- Modified test files to use `PineconeIndex` instead of `PineconeLocalIndex`
- Updated GitHub Actions workflow to simplify Python version testing
- Improved error handling and host URL management in Pinecone index methods

refactor: linter + formatter

- Changed default dimensions from 1536 to 3 in test_sync.py
- Simplified index initialization parameter for testing purposes

- Updated default dimensions to 1536 in test_sync.py
- Aligns with typical embedding model dimension requirements

- Replaced all references to PineconeIndex with PineconeLocalIndex in test files
- Added dimension handling for OpenAI and Cohere encoders in index initialization
- Updated GitHub Actions workflow to cache Poetry dependencies
- Maintained existing test logic and retry mechanisms

- Added PineconeLocalIndex to support local Pinecone testing
- Updated GitHub Actions workflow to use Pinecone local service
- Updated dependencies in pyproject.toml and poetry.lock
- Modified schema and router to support PineconeLocalIndex
- Updated test configurations to use PineconeLocalIndex