Support running the Spanner source connector without a Kafka broker

## Feature request or enhancement

_For feature requests or enhancements, please provide the following information:_

**Which use case/requirement will be addressed by the proposed feature?**

While most Debezium source connectors can run in the embedded engine / Debezium Server without a Kafka broker, the Spanner source connector currently requires one. It uses Kafka internally for cross task coordination and low watermark aggregation which makes it hard to use in broker-less deployments.

We've been exploring making it runnable without a broker for the single task case, as with one task there's no cross task state to share, so neither the coordination nor the low watermark aggregation needs a broker.

The approach pulls the Kafka-coupled coordination behind a small SPI. This keeps the existing Kafka path as the default (multi-task, unchanged) and adds a broker-less single task implementation (as a nice side effect this turns coordination into a pluggable extension point, so others could provide their own backend instead of Kafka). The mode is chosen by a simple config flag.

We have a working PoC but before going further I wanted to ask: would the maintainers be open to upstreaming something like this?

Happy to share the design, the SPI, and more details around the same.

Start by reviewing the proposed SPI, broker-less single-task mode, and configuration flag described in the issue; no repository files or tests are named. Clarify how coordination and low-watermark aggregation should work without Kafka while preserving the existing multi-task path. Done means there is an agreed upstreamable design and appropriate implementation tests.

Written by the indexing model from the issue text.