Allow DataFusion type casts for Python inputs

# What
Allow DataFusion to automatically insert type casts for Python inputs.

# Why
Currently we emit [`TypeSignature::Exact`](https://docs.rs/datafusion/latest/datafusion/logical_expr/enum.TypeSignature.html#variant.Exact) which requires users to pass exactly the types we need. That's a bit annoying for integers though, since we treat Python's `int` as unsigned and users would have to explicitly cast the inputs for unsigned data. We should probably emit [`TypeSignature::Coercible`](https://docs.rs/datafusion/latest/datafusion/logical_expr/enum.TypeSignature.html#variant.Coercible) instead.

# How
This is likely gonna interact with #161. Also we need to check if the variants of [`TypeSignatureClass`](https://docs.rs/datafusion/latest/datafusion/logical_expr/enum.TypeSignatureClass.html) are sufficient or if we need to implement a WIT bridge for [`LogicalType`](https://docs.rs/datafusion/latest/datafusion/common/types/trait.LogicalType.html) as well.

Start by reviewing the current use of TypeSignature::Exact and the related TypeSignatureClass handling, then read issue #161 for interaction details. Check whether the existing TypeSignatureClass variants can represent Python-input coercion or whether a WIT bridge for LogicalType is needed. Done means DataFusion can automatically coerce the relevant Python inputs without requiring explicit casts.

Written by the indexing model from the issue text.