Add dedicated OData API support (extract & publish)

## Summary

apiops-cli currently handles OData APIs via the generic OpenAPI fallback path, which does not capture OData-specific artifacts. This issue tracks adding first-class OData support to both the extractor and publisher.

## Background

Azure API Management added OData API support, but tooling across the ecosystem has lagged behind (see [Azure/apiops#419](https://github.com/Azure/apiops/issues/419)). Users report that entity sets and functions are lost when extracting/publishing OData APIs.

## Current behavior

- **Extractor** (`api-extractor.ts`): No special-case handling for `odata` API type — it falls through to the generic `openapi-link` export, which misses OData-specific resources.
- **Publisher** (`api-publisher.ts`): Recognizes `odata` as a valid `apiType` string but has no OData-specific import format or schema handling.
- **APIM Client** (`apim-client.ts`): `getExportFormat()` returns `'openapi-link'` for OData via the `default` case — no dedicated export format.

## What needs to change

1. **Extract**: Add OData-specific handling in `api-extractor.ts` to export CSDL schemas, entity sets, and functions via the appropriate APIM export format.
2. **APIM Client**: Add an `'odata'` case to `getExportFormat()` returning the correct APIM export format for OData APIs (investigate whether APIM supports a dedicated OData/CSDL export link).
3. **Publish**: Add OData-specific import format handling in `getImportFormat()` so that CSDL schemas and entity sets are correctly imported on publish.
4. **Tests**: Add unit/integration tests covering OData extract and publish flows.

## Related

- [Azure/apiops#419](https://github.com/Azure/apiops/issues/419) — original feature request in the apiops toolkit repo

Start in api-extractor.ts, api-publisher.ts, and apim-client.ts, tracing the existing apiType handling and getExportFormat()/getImportFormat() paths. Check the APIM API format documentation or existing client behavior to confirm the OData/CSDL export and import formats. Done means OData entity sets and functions survive extract and publish, with unit or integration tests covering both flows.

Written by the indexing model from the issue text.