fix: support case-sensitive Fabric Warehouse collations

[sdebruyn]

Summary

Fixes #373.

FabricAdapter does not override _make_match_kwargs, so dbt-adapters' default runs unchanged and lowercases identifiers whenever quoting.case_sensitive is False (the dbt-core default). On a Fabric Warehouse provisioned with a case-sensitive collation (e.g. SQL_Latin1_General_CP1_CS_AS), dbt then asks Fabric for a relation by its lowercased name and Fabric correctly answers "no such object" — making the adapter unusable on that warehouse.

How

Add a one-method override on FabricAdapter that uses filter_null_values to drop Nones but preserves the original casing of database, schema, and identifier:

def _make_match_kwargs(self, database: str, schema: str, identifier: str) -> Dict[str, str]:
    return filter_null_values(
        {
            "database": database,
            "identifier": identifier,
            "schema": schema,
        }
    )

Behaviour on case-insensitive collations is unchanged: Fabric's identifier comparison ignores case on those collations regardless of what we pass in.

Test plan

• Unskip TestCachingUppercaseModel (was marked pytest.mark.skip(reason="Fabric DW does not support Case Insensivity.")). It exercises the relation-cache hit path with mixed-case identifiers; with the override in place it now passes against case-sensitive collations and continues to pass against case-insensitive ones.
• I'm a non-Microsoft contributor and cannot run the CI against the Microsoft Fabric tenant. Happy to add a local test-run screenshot if a maintainer wants to engage with the change.

Closes #373