Initial Parquet File/Dataset abstraction

[kylebarron]

Prep work for being able to fetch metadata from a bunch of Parquet files in a folder and load a spatial filter

Change list

• New ParquetFile and ParquetDataset Rust structs to read from one or multiple Parquet files. This is generic over AsyncFileReader, which primarily works with object store.
• Added initial metadata handling, to e.g. read the bounding box of the file.
• Added Python bindings to each class. This uses ObjectStore.
• Added initial JS bindings to each class. This uses a custom implementation of AsyncFileReader vendored from parquet-wasm. I couldn't get the ObjectStore integration working just yet. Thought object_store_wasm_s3 was updated and that might be an interesting thing to check out.

[kylebarron]

@H-Plus-Time I'm starting to explore an object-store based async JS Parquet reader in this branch. I'm coming to think it's the best solution because I can reuse so much of the same Rust and Python code.

I'm thinking of trying to implement a minimal impl based on ehttp for the web, but very welcome to advice/thoughts.

[kylebarron]

I failed at trying to implement a wasm ObjectStore. In particular I ended up hitting Send/Sync issues, and since ObjectStore is defined as #[async_trait] and not #[async_trait(?Send)], I'm not sure it's even possible to implement ObjectStore in wasm32. Instead, for now I vendored the AsyncFileReader impl into here.